The right prompt.
The right action.

EXZA Remote — detailed execution plan: Dimensioned 3D cube OBJECTIVE Check unit scale and specified dimensions; propose Blender Python to construct and verify bounding box. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check unit scale and specified dimensions; propose Blender Python to construct and verify bounding box. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Business introduction email OBJECTIVE Prepare a concise subject and introduction for recipient and service; do not invent testimonials or send automatically. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Prepare a concise subject and introduction for recipient and service; do not invent testimonials or send automatically. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Repeatable test pipeline OBJECTIVE Plan exact test commands, side effects, dependencies and success criteria; ask before running. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Plan exact test commands, side effects, dependencies and success criteria; ask before running. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Android Gradle failure review OBJECTIVE Correlate actual stack traces with JDK, SDK and Gradle versions; propose smallest isolated fix. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Correlate actual stack traces with JDK, SDK and Gradle versions; propose smallest isolated fix. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Structure meeting minutes OBJECTIVE Extract decisions, assignments and verified deadlines from authorized notes; do not invent names. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Extract decisions, assignments and verified deadlines from authorized notes; do not invent names. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Compare CSV headers OBJECTIVE Verify delimiters, data types and missing columns in bounded approved input. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Verify delimiters, data types and missing columns in bounded approved input. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Electronics enclosure OBJECTIVE Request PCB dimensions, wall thickness and screw spacing; make separate lid and base; check clearances. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Request PCB dimensions, wall thickness and screw spacing; make separate lid and base; check clearances. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Answer invoice question OBJECTIVE Verify provided invoice number, payment record and date; draft a courteous response without assuming anything unpaid. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Verify provided invoice number, payment record and date; draft a courteous response without assuming anything unpaid. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Monitor an EXZA test job OBJECTIVE Inspect a real authorized job ID and bounded output; do not invent background notifications. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect a real authorized job ID and bounded output; do not invent background notifications. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: JavaScript stack trace explanation OBJECTIVE Find referenced sources and package versions, distinguish proven cause and hypothesis. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Find referenced sources and package versions, distinguish proven cause and hypothesis. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Draft a professional quotation OBJECTIVE Use supplied quantities and prices, flag tax assumptions and avoid invented terms. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use supplied quantities and prices, flag tax assumptions and avoid invented terms. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Find duplicate row identifiers OBJECTIVE Check unique IDs within readable portions, flag partial coverage and protect private records. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check unique IDs within readable portions, flag partial coverage and protect private records. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Printable hinge OBJECTIVE Specify pin diameter, rotational clearance, print direction and manifold geometry; verify interference. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Specify pin diameter, rotational clearance, print direction and manifold geometry; verify interference. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Attachment-ready email review OBJECTIVE Check recipient and attachment file metadata; explain that EXZA Mail tools may not support adding attachments. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check recipient and attachment file metadata; explain that EXZA Mail tools may not support adding attachments. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Safe batch filename change OBJECTIVE Map source and destination names, check conflicts and rollback; get item-level approval. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Map source and destination names, check conflicts and rollback; get item-level approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: API error-handling audit OBJECTIVE Review input validation, exception handling, error information leakage and bounded responses. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review input validation, exception handling, error information leakage and bounded responses. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Prepare project handover document OBJECTIVE Summarize verified status, paths, dependencies and remaining work, no invented milestones. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Summarize verified status, paths, dependencies and remaining work, no invented milestones. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Summarize expense categories OBJECTIVE Check currency, signed amounts, dates and category mappings; avoid claiming sampled rows represent totals. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check currency, signed amounts, dates and category mappings; avoid claiming sampled rows represent totals. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Servo bracket OBJECTIVE Require actual servo datasheet and measured mounting holes; model with test tolerance; export provisional STL. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Require actual servo datasheet and measured mounting holes; model with test tolerance; export provisional STL. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Polite payment reminder OBJECTIVE Use only verified due dates and received payments; avoid threats and request approval before any send. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use only verified due dates and received payments; avoid threats and request approval before any send. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Monthly report folder planning OBJECTIVE Review existing directory names, propose dated structure and create only after approval. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review existing directory names, propose dated structure and create only after approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Filename path-safety review OBJECTIVE Inspect normalized allowed roots, symlink handling and no-overwrite protections without probing production. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect normalized allowed roots, symlink handling and no-overwrite protections without probing production. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Create repeatable SOP OBJECTIVE Include prerequisites, step-by-step approvals, checks, exceptions and rollback; review before saving. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Include prerequisites, step-by-step approvals, checks, exceptions and rollback; review before saving. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Compare inventory snapshots OBJECTIVE Align stable product IDs, quantify added and removed items, handle duplicate IDs and missing quantities. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Align stable product IDs, quantify added and removed items, handle duplicate IDs and missing quantities. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Low-poly game props OBJECTIVE Define triangle budget, origin and asset sizes; create distinct meshes and material variants. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Define triangle budget, origin and asset sizes; create distinct meshes and material variants. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Ask for bug reproduction steps OBJECTIVE Ask the tester for OS, app version, redacted error and repeatable steps; never ask for passwords or PIN. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Ask the tester for OS, app version, redacted error and repeatable steps; never ask for passwords or PIN. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Validate and zip build outputs OBJECTIVE Verify expected files and hashes, show exact archive command and check output after approval. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Verify expected files and hashes, show exact archive command and check output after approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Python dependency troubleshooting OBJECTIVE Compare version constraints and interpreter, propose isolated verification, no global installs. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Compare version constraints and interpreter, propose isolated verification, no global installs. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Write a support knowledge article OBJECTIVE Document real symptoms, affected OS, safe diagnostics and escalation without collecting secrets. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Document real symptoms, affected OS, safe diagnostics and escalation without collecting secrets. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Inspect JSON export schema OBJECTIVE Validate accessible structure, nested keys and type mismatches, mask secrets and preserve source. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Validate accessible structure, nested keys and type mismatches, mask secrets and preserve source. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Atmospheric horror scene OBJECTIVE Block out terrain and props, set dramatic lighting and volumetrics, preview low-resolution composition. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Block out terrain and props, set dramatic lighting and volumetrics, preview low-resolution composition. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Bilingual release announcement OBJECTIVE Check actual release facts, maintain equal meaning across English/German and preview both drafts before sending. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check actual release facts, maintain equal meaning across English/German and preview both drafts before sending. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Summarize errors from logs OBJECTIVE Bound search by approved log directory and error code, redact sensitive values and note sampling limits. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Bound search by approved log directory and error code, redact sensitive values and note sampling limits. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Database migration preflight OBJECTIVE Review schema, ordering, rollback and fixtures; never run destructive SQL automatically. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review schema, ordering, rollback and fixtures; never run destructive SQL automatically. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Hard-surface bevels OBJECTIVE Inspect applied scale and normals; add reversible bevel modifier; confirm target dimensions remain correct. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect applied scale and normals; add reversible bevel modifier; confirm target dimensions remain correct. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Meeting request with agenda OBJECTIVE Write a meeting request with explicit timezone, three agenda points and optional slots without pretending calendar access. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Write a meeting request with explicit timezone, three agenda points and optional slots without pretending calendar access. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: PowerShell script safety review OBJECTIVE Inspect external calls, paths, elevation and file deletion before execution, no automatic run. TARGET the authorized Windows PC. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect external calls, paths, elevation and file deletion before execution, no automatic run. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Responsive web component review OBJECTIVE Read affected CSS and markup, propose minimal changes and viewport regression checks. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Read affected CSS and markup, propose minimal changes and viewport regression checks. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Brushed-metal material OBJECTIVE Build version-compatible shader nodes, roughness variation and directional texture; validate preview. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Build version-compatible shader nodes, roughness variation and directional texture; validate preview. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Supplier quotation request OBJECTIVE List exact part dimensions, quantity and delivery question; claim drawings attached only if verified. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN List exact part dimensions, quantity and delivery question; claim drawings attached only if verified. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Bash job safety review OBJECTIVE Review quoting, globbing, sudo, rm and output paths; propose dry run but do not execute. TARGET the authorized Linux machine. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review quoting, globbing, sudo, rm and output paths; propose dry run but do not execute. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Investigate flaky tests OBJECTIVE Compare passing and failing logs, timing and shared resources, avoid silent skip or larger timeout without evidence. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Compare passing and failing logs, timing and shared resources, avoid silent skip or larger timeout without evidence. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Studio product lighting OBJECTIVE Plan camera lens, background and key, fill and rim lights; test exposure and clean shadow. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Plan camera lens, background and key, fill and rim lights; test exposure and clean shadow. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Sensitive support-mail screening OBJECTIVE Review recipient, pasted message and possible credential disclosure; redact data and do not send. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review recipient, pasted message and possible credential disclosure; redact data and do not send. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Automate Blender render preparation OBJECTIVE Check Blender binary and selected file, output naming and preview, then request approved monitored execution. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check Blender binary and selected file, output naming and preview, then request approved monitored execution. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Design safe CLI help text OBJECTIVE Document verified flags, indicate writes and approvals, never invent options not in source. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Document verified flags, indicate writes and approvals, never invent options not in source. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Cycles render optimization OBJECTIVE Check installed render engine and CPU/GPU support; tune samples, bounces and denoise with benchmark frames. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check installed render engine and CPU/GPU support; tune samples, bounces and denoise with benchmark frames. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Service cancellation email OBJECTIVE Use confirmed contract reference and cancellation date; request written confirmation; avoid false legal claims. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use confirmed contract reference and cancellation date; request written confirmation; avoid false legal claims. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Checksum manifest comparison OBJECTIVE Verify supported hashing tool and compare approved source and copy without overwriting. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Verify supported hashing tool and compare approved source and copy without overwriting. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Refactor source module safely OBJECTIVE Map callers and current tests, propose small diff and verification, no automatic edits. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Map callers and current tests, propose small diff and verification, no automatic edits. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Eevee preview pipeline OBJECTIVE Check Blender version, select supported Eevee engine, lower render scale, preserve production settings. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check Blender version, select supported Eevee engine, lower render scale, preserve production settings. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Approve final outgoing email OBJECTIVE Read back full address, subject and body, explain irreversible sending, ask for explicit approval. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Read back full address, subject and body, explain irreversible sending, ask for explicit approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Draft changelog from Git OBJECTIVE Read approved diff and test results; distinguish completed fixes from proposed changes. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Read approved diff and test results; distinguish completed fixes from proposed changes. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Localization completeness check OBJECTIVE Compare approved German and English key sets, note missing keys and context without guessing legal translations. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Compare approved German and English key sets, note missing keys and context without guessing legal translations. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Camera orbit animation OBJECTIVE Define fps, pivot and orbit length; generate keyframes and validate smooth interpolation and last frame. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Define fps, pivot and orbit length; generate keyframes and validate smooth interpolation and last frame. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Email draft without mailbox access OBJECTIVE Write copy-ready subject and body in ChatGPT; do not claim EXZA can send on Windows/Linux. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Write copy-ready subject and body in ChatGPT; do not claim EXZA can send on Windows/Linux. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Preview folder cleanup rules OBJECTIVE List exact potential moves and collisions; require consent and use recoverable trash only if chosen. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN List exact potential moves and collisions; require consent and use recoverable trash only if chosen. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Desktop installer artifact review OBJECTIVE Inspect manifests, architecture and required assets; test packaging without promising signatures. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect manifests, architecture and required assets; test packaging without promising signatures. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Product turntable animation OBJECTIVE Choose object or camera rotation, fix pivot, keep lighting stable and verify loop continuity. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Choose object or camera rotation, fix pivot, keep lighting stable and verify loop continuity. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Short project status email OBJECTIVE Summarize verified work, blocker and next milestone in under 120 words, without sending. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Summarize verified work, blocker and next milestone in under 120 words, without sending. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Cross-platform prerequisites checklist OBJECTIVE Identify actual OS and supported tools, propose version-only checks and mark unsupported commands. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Identify actual OS and supported tools, propose version-only checks and mark unsupported commands. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: ESP32 project pin review OBJECTIVE Read board-specific pin definitions and dependencies, flag hardware uncertainty, compile before flashing. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Read board-specific pin definitions and dependencies, flag hardware uncertainty, compile before flashing. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: GLB export for web OBJECTIVE Validate mesh scale, material compatibility, texture paths and triangle budget; inspect exported file. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Validate mesh scale, material compatibility, texture paths and triangle budget; inspect exported file. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Apology for a delayed response OBJECTIVE Acknowledge delay, give realistic next step and omit invented deadlines; draft only. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Acknowledge delay, give realistic next step and omit invented deadlines; draft only. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Stop a monitored EXZA job OBJECTIVE Identify EXZA-owned job and impact; request confirmation before using the job-stop tool. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Identify EXZA-owned job and impact; request confirmation before using the job-stop tool. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Plan an application rollback OBJECTIVE Identify known artifacts, database compatibility and approval checkpoints without deploying. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Identify known artifacts, database compatibility and approval checkpoints without deploying. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Watertight STL export OBJECTIVE Check manifold edges, normals, disconnected shells and millimeter scale; advise slicer test. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check manifold edges, normals, disconnected shells and millimeter scale; advise slicer test. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Summarize pasted email thread OBJECTIVE Use only approved pasted thread, extract open questions, redact personal details and propose a reply. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use only approved pasted thread, extract open questions, redact personal details and propose a reply. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Recurring script plan without scheduler OBJECTIVE Draft idempotent script with inputs, output paths and dry-run; do not claim unattended scheduling is enabled. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Automated scheduling or unattended recurring execution is not granted by this prompt. A monitored terminal job requires the actual supported EXZA tool and approval. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Draft idempotent script with inputs, output paths and dry-run; do not claim unattended scheduling is enabled. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Create feature test checklist OBJECTIVE Map positive, negative and cross-platform cases using actual source, propose commands before execution. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. Use only approved source files and supported EXZA tools; do not assume third-party app-specific integrations are available. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Map positive, negative and cross-platform cases using actual source, propose commands before execution. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Scene collection cleanup OBJECTIVE Audit object and collection names, propose naming rules, preserve parents and animation references. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Audit object and collection names, propose naming rules, preserve parents and animation references. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Confirm a customer appointment OBJECTIVE Draft a clear confirmation with verified date, time, timezone and meeting format; confirm recipient before sending. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Draft a clear confirmation with verified date, time, timezone and meeting format; confirm recipient before sending. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Repair mesh normals OBJECTIVE Inspect face orientation, custom normals and smoothing; repair only in a copy and compare results. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect face orientation, custom normals and smoothing; repair only in a copy and compare results. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Respond to refund inquiry OBJECTIVE Draft a factual answer using confirmed policy and transaction facts; do not promise refund before approval. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Draft a factual answer using confirmed policy and transaction facts; do not promise refund before approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Geometry Nodes scatter OBJECTIVE Confirm API version and density; expose random seed and scale; preview with low instance count. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Confirm API version and density; expose random seed and scale; preview with low instance count. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Request a missing document OBJECTIVE Ask for a specific missing document using privacy-conscious transmission guidance and no unnecessary data. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Ask for a specific missing document using privacy-conscious transmission guidance and no unnecessary data. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Parametric drilled plate OBJECTIVE Set thickness, hole diameter, center distance and edge margin; check clearances and dimensions. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Set thickness, hole diameter, center distance and edge margin; check clearances and dimensions. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Check email wording for legal claims OBJECTIVE Flag unsupported legal terms and deadlines in supplied draft, propose cautious wording and suggest professional review. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Flag unsupported legal terms and deadlines in supplied draft, propose cautious wording and suggest professional review. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: UV unwrap and texture checker OBJECTIVE Inspect seams, island margins and distortion; display checker texture before texture painting. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect seams, island margins and distortion; display checker texture before texture painting. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Follow up after software demo OBJECTIVE Summarize agreed next steps and ask for feedback without inventing customer interest or commitments. TARGET the authorized Mac. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. EXZA macOS Mail drafting/sending tools are Mac-only; drafts for Windows/Linux can be created in chat, but do not claim an actual send. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Summarize agreed next steps and ask for feedback without inventing customer interest or commitments. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Game mesh decimation OBJECTIVE Measure triangles, set a target budget, compare silhouette and normals against original. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Measure triangles, set a target budget, compare silhouette and normals against original. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Set pivots and transforms OBJECTIVE Inspect object hierarchy and animation dependencies before changing origins or applying scale. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect object hierarchy and animation dependencies before changing origins or applying scale. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Basic mechanical rig OBJECTIVE Define joint axes, rotation limits and hierarchy; test movements for collision and distortion. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Define joint axes, rotation limits and hierarchy; test movements for collision and distortion. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Emboss 3D labels OBJECTIVE Check font licensing, stroke width, extrusion depth and print readability; preview placement. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check font licensing, stroke width, extrusion depth and print readability; preview placement. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Mechanical assembly fit check OBJECTIVE Align parts using measured references; inspect collisions, screw access and clearance; do not assert physical fit. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Align parts using measured references; inspect collisions, screw access and clearance; do not assert physical fit. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Transparent PNG product render OBJECTIVE Configure RGBA and film transparency, check shadows and alpha using a low-resolution preview. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Configure RGBA and film transparency, check shadows and alpha using a low-resolution preview. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Batch rendering scene variants OBJECTIVE Check Blender CLI and render engine, estimate output size and show the exact batch command before approval. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check Blender CLI and render engine, estimate output size and show the exact batch command before approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Fix missing external textures OBJECTIVE Inventory referenced textures within allowed paths; propose relative path repairs in a duplicate scene. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inventory referenced textures within allowed paths; propose relative path repairs in a duplicate scene. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Compare two BLEND versions OBJECTIVE Use approved Blender Python metadata inspection for objects, mesh counts, materials and dimensions. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use approved Blender Python metadata inspection for objects, mesh counts, materials and dimensions. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Prepare Godot GLB assets OBJECTIVE Check axis conventions, scale, origins and materials; export a copy and suggest import validation. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Check axis conventions, scale, origins and materials; export a copy and suggest import validation. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Prepare Unreal asset export OBJECTIVE Confirm Unreal import requirements, centimeters/meter conversion, collision naming and supported format. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Confirm Unreal import requirements, centimeters/meter conversion, collision naming and supported format. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: 3D printer tolerance coupon OBJECTIVE Model marked holes, pegs and slots with controlled tolerance increments; check watertightness before printing. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Model marked holes, pegs and slots with controlled tolerance increments; check watertightness before printing. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Reusable material swatch library OBJECTIVE Define PBR roughness and color variants; assign to test objects and compare renders consistently. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Define PBR roughness and color variants; assign to test objects and compare renders consistently. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Pack scene dependencies OBJECTIVE Inventory relative texture paths and external files; plan packing or copying only after approval. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inventory relative texture paths and external files; plan packing or copying only after approval. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Create a modular wall kit OBJECTIVE Specify grid dimensions, snap interfaces and texture scale; validate pieces align on a sample grid. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Specify grid dimensions, snap interfaces and texture scale; validate pieces align on a sample grid. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Build a simple staircase OBJECTIVE Calculate riser and tread dimensions from total rise and step count; generate measured stair geometry. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Calculate riser and tread dimensions from total rise and step count; generate measured stair geometry. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Create cable routing channel OBJECTIVE Ask for cable diameter, bend radius, wall thickness and mounting route; validate print clearances. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Ask for cable diameter, bend radius, wall thickness and mounting route; validate print clearances. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Render a clay concept preview OBJECTIVE Use a neutral clay material, soft studio light and consistent camera to inspect silhouette without costly materials. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Use a neutral clay material, soft studio light and consistent camera to inspect silhouette without costly materials. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Create a photogrammetry cleanup plan OBJECTIVE Inspect holes, noisy surfaces, scale reference and mesh density; propose non-destructive cleanup steps. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Inspect holes, noisy surfaces, scale reference and mesh density; propose non-destructive cleanup steps. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — detailed execution plan: Blender Python script code review OBJECTIVE Review bpy script for destructive operators, version mismatches, path writes and missing output validation before running. TARGET the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools. 1. PREFLIGHT AND ACCESS Confirm which device is actually paired, its OS, owner-approved scope and the real EXZA tool descriptions. For Blender tasks, check whether Blender and a compatible bpy/Python interface are installed. EXZA can inspect permitted files and propose scripts; do not claim it can directly manipulate Blender mesh vertices by itself. For scripted changes, show a version-compatible Blender background invocation and exact output paths, then require owner approval to run it. GUI keyboard automation is limited and may not be reliable for precision modeling. A model or render exists only if actual result files can be verified. 2. TARGETED DISCOVERY Ask me for required dimensions, versions, exact object names, file paths, file formats or recipients whenever they are needed. Inspect only the selected, authorized resources with bounded read-only tools. Check existing files, relevant documentation, file size limits and possible conflicts before taking action. 3. TASK-SPECIFIC PLAN Review bpy script for destructive operators, version mismatches, path writes and missing output validation before running. Propose concrete ordered steps, expected artifacts, meaningful filenames, measurable acceptance criteria and a small preview or test stage. Explain likely failure modes and how to preserve the original state. 4. APPROVAL BEFORE MODIFICATION Show the exact file edit, Blender Python script, terminal command, GUI text, or email recipient and message as applicable. Explain permissions, data exposure, possible side effects, target paths, estimated output size if relevant, and any non-reversible changes. Do not interpret this sample prompt as authorization. Ask for explicit approval through ChatGPT and local EXZA owner controls; never bypass denials or elevate privileges. 5. IMPLEMENTATION AND VALIDATION Only after approval, use a genuinely supported EXZA operation within the authorized workspace. Never overwrite original assets, delete files, mail recipients or trigger external actions implicitly. Confirm real tool output and independently verify results using read-only file metadata, project checks, rendered preview or test command where supported. If a step is unavailable, stop that step honestly. 6. REPORT Provide: preflight, evidence, proposed changes, approvals needed, executed actions, verified outputs, remaining limitations and precise next steps. State which operations were not performed. Mask credentials and unrelated personal data.

EXZA Remote — practical field guide for Diagnose a Windows path quoting issue GOAL Examine why [COMMAND] fails when a Windows project path contains spaces or special characters. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Inspect existing error, distinguish PowerShell from cmd quoting and propose harmless dry-run examples before any execution. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review Windows project archives GOAL Inspect the names and sizes of [PROJECT BACKUPS] without unpacking or modifying archives. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Verify archive filenames, dates and apparent version labels in authorized folders; show naming inconsistencies. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Windows pairing readiness GOAL Help me prepare [WINDOWS PC] for an EXZA Remote connection without making system changes. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check the actual available installer instructions, local owner confirmation and network reachability; distinguish unverified Windows functionality. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Summarize an accessible issue log GOAL Summarize repeated issues in approved [ISSUE LOG] without assuming they are all still open. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Extract dates, observed failures, fixes and unverified status from bounded file sections; avoid exposing personal identifiers. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Trace a user-facing error message GOAL Locate the exact user-visible error [MESSAGE] in approved project sources and identify the surrounding feature. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search literal text first, then relevant localization keys within bounded directories; report references and do not edit. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review a proposed backup deletion GOAL Assess risks of removing [BACKUP FILE] while keeping a recoverable option wherever EXZA permits. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check file name, destination, other backup copies and explicit approval; prefer EXZA recoverable trash; never permanently delete silently. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Check macOS app launch prerequisites GOAL Determine whether approved [APP NAME] can be launched by the available EXZA Mac desktop tools. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence List installed apps and current windows, compare exact app name, then ask before initiating a launch. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Draft a customer setup follow-up GOAL Prepare a Mac Mail follow-up to [CUSTOMER] about an incomplete EXZA setup without assuming the cause. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Ask for the exact observed error, avoid asking for passwords or PINs and provide safe troubleshooting steps; send only after confirmation. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Inspect a Linux permission-denied build GOAL Investigate a permission-denied error in an authorized [PROJECT] build without raising privileges. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Read available error and path metadata, identify ownership hypothesis versus confirmed restrictions, avoid sudo, chmod or chown changes. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Check Linux permission metadata GOAL Inspect metadata for [PROJECT FILES] inside the permitted Linux workspace. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use EXZA metadata and supported read-only shell inspection only if authorized; do not run chmod or change ownership. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Linux pairing readiness GOAL Create a step-by-step readiness checklist for linking [LINUX DEVICE] to EXZA Remote. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use documented launcher and pairing flow, explain Y/N owner decision, do not assume root access or install packages. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Prepare a deployment runbook draft GOAL Draft a repeatable release runbook for [PROJECT] from approved documentation without executing deployment. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Include prerequisites, backup, dry-run, approval gates, verification, failure modes and rollback only where backed by source. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Locate image references in web source GOAL Find references to [IMAGE FILE] in authorized HTML/CSS/JS without opening private assets. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search filename and related paths, group direct links from CSS and HTML and note dynamic references as uncertain. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Design minimal project access for EXZA GOAL Plan a least-privilege scope for EXZA access to [PROJECT] and its required tasks. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Identify read-only actions, exact permitted folders and separate approval needs for editing or running commands; do not alter security settings automatically. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Prepare safe multi-step UI input GOAL Plan multiple text entries in [APP] without accidentally submitting or affecting a different window. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Confirm focus at each step, preview the exact text and shortcuts, request approval and stop on unexpected dialogs. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review a release announcement for accuracy GOAL Prepare [RELEASE EMAIL] and verify new-feature claims against approved release documentation before sending. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Separate confirmed features, limitations and platform-specific notes; review final recipient and body and do not send automatically. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review macOS build toolchain GOAL Inspect [PROJECT] requirements and safely verify the installed developer toolchain on the Mac. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use approved version-only commands for tools named by the README, avoid installing Xcode components automatically. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review Mac screenshots folder GOAL Inventory screenshot filenames in my authorized Pictures folder without reading image content. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use a bounded file listing and metadata; group by date and name patterns; no deletion or upload. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Investigate a Mac connection drop GOAL Investigate why my paired Mac occasionally becomes disconnected without altering connectivity. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Collect EXZA status, recent authorized connection events and actual failure messages; distinguish pause, revoke, exit and network outage. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Draft an onboarding checklist GOAL Create a beginner-friendly onboarding checklist for [PROJECT] based on its approved README. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Separate OS-specific commands, safety approvals and prerequisites; mark unknown steps and do not install software. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Find Windows installer references GOAL Find which approved project files refer to its Windows installer [INSTALLER NAME]. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Inspect read-only filenames and safe text references; never execute installers or print embedded credentials. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review remote email privacy risks GOAL Evaluate an EXZA-assisted email workflow for accidental disclosure of private file content or addresses. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check recipient, attachment scope and included details, prefer preview and redaction and demand final send confirmation. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Inspect Mac apps for a project workflow GOAL Identify which already installed Mac applications might be useful for [WORKFLOW], without launching them. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use EXZA installed-app list, match names to the task and distinguish likely relevance from confirmed functionality. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Draft a helpful reply to [QUESTION] with subject and body; never open Mail or send without separate approval. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Prepare a safe rollback command review GOAL Assess the exact proposed [ROLLBACK COMMAND] for [PROJECT] before anyone runs it. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Identify affected paths, destructive flags and existing backups; show a safer preview or dry-run if supported and await explicit approval. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Prepare a versioned backup naming plan GOAL Propose consistent safe names for copies of [PROJECT] before the next release. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Inspect existing filenames and versions; avoid overwrites and distinguish backup copies from production sources. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review paired-device permissions GOAL Compare the permissions and status of selected EXZA devices while protecting private device details. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Name the currently selected device, inspect only exposed scope and approval mode, never silently switch another owner account. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Analyze two CSV exports for drift GOAL Compare permitted CSV snapshots [OLDER CSV] and [NEWER CSV] within file limits. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Verify common column names, unique row identifiers and data types; describe only the sampled portion if truncated. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Trace Linux launcher documentation GOAL Locate the instructions and scripts for [LINUX LAUNCHER] in an authorized repository. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search known script names then README references; distinguish documented actions from tested operation. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Explain the EXZA PIN and physical Y/N approval flow and which computer operations remain blocked without owner consent. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review a Mac clipboard paste workflow GOAL Plan copying prepared [TEXT] and pasting into selected [APP] without reading unrelated clipboard data. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Confirm app window, show new text, explain clipboard overwrite and request separate write and paste permissions. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Prepare a service notice for [AUDIENCE] using only confirmed scope and dates; show short and full variants without sending. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Inspect Windows Node project scripts GOAL Review npm scripts of [NODE PROJECT] on Windows and distinguish safe checks from commands that modify files. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Read package.json, classify each script, flag pre/post hooks and do not run install, build or clean without permission. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review a large file by chunks GOAL Analyze [LARGE TEXT FILE] using bounded EXZA file chunks with a final coverage report. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Verify file version and checksum at each continuation; respect offsets and file-size caps, avoid secrets and claim only read sections. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Safe troubleshooting decision tree GOAL Create a diagnostic flow for [EXZA ERROR] with evidence-based decisions and no permission bypass. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Separate offline agent, denied scope, expired OAuth, unpaired device and unsupported action; propose only official recovery paths. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Draft a user-facing troubleshooting guide GOAL Write a customer help article for [KNOWN ERROR] using evidence from approved error messages and documentation. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Give symptoms, affected platforms, safe checks, official recovery and escalation; do not recommend bypassing approval prompts. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Review unused-looking files carefully GOAL Find files whose names suggest old [FEATURE] versions, without automatically classifying them as unused. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check references within accessible source and mark uncertainty when search limits or dynamic imports prevent a complete answer. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Assess [ACTION] for read, write, terminal and desktop permissions and show safer alternatives without executing it. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: List visible application windows on the authorized Mac without focusing apps or reading their private content. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Review recipient [ADDRESS], subject [SUBJECT] and message [BODY] before irreversible sending on macOS Mail; require explicit approval. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review a Linux container configuration GOAL Review authorized Docker Compose or container definitions for [SERVICE] without launching containers. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Read safe source files only, identify ports, mounts and environment-variable names; mask values and avoid docker execution. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Check a folder against a manifest GOAL Verify whether [FOLDER] contains each named file from [MANIFEST] without creating or deleting anything. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Normalize filenames carefully, compare exact names and note unreadable entries rather than guessing they are absent. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Check the authorized Windows PC for OS, architecture, CPU, RAM and uptime using the available read-only EXZA system information. Keep unknown values unknown. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review technical release notes GOAL Verify claims in [RELEASE NOTES] against approved source and test artifacts before publishing. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Match each claimed feature to documented code and test results; flag anything unverified instead of rewriting facts. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Find all migration notes GOAL Find the approved notes and scripts mentioning [MIGRATION NAME] and construct a chronology. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search names and terms, inspect only selected small text files, sort by recorded dates rather than assumed chronology. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Check available connection state and explain local terminal stop, web disconnect and device removal without bypassing controls. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Check that [APP NAME] is installed, then request permission before launching it on the Mac. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Draft a follow-up on [REQUEST] based only on provided dates and facts; show the text before any Mail draft is created. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Analyze Windows test runner output GOAL Triage failure output of the approved Windows [TEST RUN]. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use existing bounded logs, separate failed assertions from skipped tests, propose a focused rerun only after approval. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Prepare a no-overwrite rename batch GOAL Plan renaming selected [FILES] to [NEW NAMING RULE] without performing the batch yet. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check each original, target collision and ordering; request item-level approval and never overwrite existing files. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Identify kernel, distribution if available, CPU, RAM and uptime for the authorized Linux device. Do not infer a distribution from the kernel alone. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Read approved [NOTES] and extract tasks, owners and due dates without inventing missing information. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review test coverage references GOAL Identify accessible tests that mention [FUNCTION NAME] and distinguish references from actual executed coverage. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search test filenames and text, relate matches to package scripts and report gaps without claiming tests passed. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Review [PROMPT] for passwords, broad folder access and unnecessary data sharing, then rewrite it to minimize permissions. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Find the exact [APP NAME] window in the visible Mac window list and request approval before switching focus. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Diagnose a Mac launch script failure GOAL Investigate why authorized [START SCRIPT] on macOS exits without showing the expected prompt. TARGET my authorized Mac. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Confirm the Mac-specific tool is available with the necessary device permission before using it. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Review shell syntax, permission metadata and actual captured output; avoid chmod and unattended relaunch until approved. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Investigate a missing project artifact GOAL Find where the expected [ARTIFACT NAME] should be inside the approved Linux project. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Compare expected build output locations from docs to bounded filename search; do not rebuild or generate files automatically. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Summarize the current Mac connection, uptime, memory and bounded process information. Distinguish observations from a performance diagnosis. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inspect only a bounded part of approved [CSV FILE], identify delimiter, columns and obvious missing values. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Find accessibility labels GOAL Locate labels and aria attributes for [UI COMPONENT] within approved frontend source. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Search related component names in HTML and JS; report visible labels and missing obvious translations only from verified source. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Verify approved [CHANGE] using read-only EXZA evidence; report success, failure or uncertainty without retrying writes automatically. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Prepare [TEXT] for the focused Mac app [APP] and request confirmation of target window and text before typing; do not press Return. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Compare Linux application configurations GOAL Compare two approved text configs [CONFIG A] and [CONFIG B] and summarize behavioral differences. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Mask credentials, compare key names and nonsecret values, do not apply new settings or restart any service. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Plan safe Windows file reconciliation GOAL Compare authorized [SOURCE FOLDER] with [DESTINATION FOLDER] by visible metadata and plan a collision-free reconciliation. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Distinguish missing, changed-size and same-name items; do not claim content equality without hashes and do not copy until approved. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Find high-memory processes using the bounded EXZA process list on my Windows machine. Do not terminate or label them malicious. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Turn [LOG] from an approved workspace into evidence-based expected/actual behavior and reproduction steps, removing personal data. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Find accessible README-like filenames under [PROJECT] using bounded EXZA search; show the paths grouped by subfolder. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Prepare [TEXT] for the Mac clipboard and ask before replacing current contents; do not read the previous clipboard. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review a failed background job safely GOAL Analyze the status and output of EXZA-owned job [JOB ID] without restarting or stopping it. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Use EXZA job status, timestamps and bounded output; distinguish running, failed and unknown results and list safe next actions. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — practical field guide for Inspect a corrupted text export GOAL Determine why [TEXT EXPORT] looks truncated or contains unexpected characters without modifying it. TARGET the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. App-window automation and macOS Mail tools are Mac-specific, not universal. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Check size, encoding markers, accessible text chunks and observed errors; propose safe recovery only after evidence. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Identify currently visible workloads and resource use on my Linux host without reading private process arguments or terminating anything. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Use the approved README and tests of [PROJECT] to map passed evidence, missing checks and rollback steps without deploying. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Search allowed source files for exact [OLD VALUE] and identify candidates for manual migration, without changing every match blindly. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Explain the possible effect of [SHORTCUT] in focused [APP] and ask before sending any keyboard event. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Plan a Windows Python venv check GOAL Inspect [PYTHON PROJECT] for documented virtual-environment expectations on Windows. TARGET my authorized Windows computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Read requirements and metadata, propose safe Python version checks; creating a venv requires explicit approval. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: List [PROJECT FOLDER], find README and entry-point filenames, then read only chosen small text files within EXZA limits. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Check the documented prerequisites for [PROJECT] on Windows; propose version checks for Node, Git and Python only where they are relevant and authorized. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Read the approved source section for [FEATURE], prepare a minimal diff and validation plan, and do not edit until approved. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare key sets of the approved German and English localization files and identify missing keys without inventing translations. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — practical field guide for Review a Linux scheduled task definition GOAL Review an approved user-owned scheduled-task definition [SCHEDULE FILE] without installing or changing it. TARGET my authorized Linux computer. STEP 1 — Verify available access Check the actual paired device, its platform, current approval scope and the available EXZA tools. Windows and Linux have not been fully verified natively; do not promise Mac-only desktop or Mail features. If a tool is absent or blocked, disclose the limitation; never improvise a bypass. STEP 2 — Gather task-specific evidence Read only allowed text definition and interpret frequency, working directory and file effects; do not modify cron or timers. Clarify remaining bracketed placeholders before reading files or invoking tools. Use the least-privileged, narrowest read-only operation that answers the question. Respect tool pagination, size limits and the approved workspace; avoid hidden files, credentials and unrelated personal documents. STEP 3 — Interpret findings Differentiate verified output from assumptions and missing observations. Explain which exact filenames, paths, versions, windows or commands support each conclusion. If a system-specific command is needed, propose the right syntax for the confirmed operating system and explain what it does before execution. STEP 4 — Propose any changes Present the exact proposed command, file edit, UI action, destination or outgoing recipient first. Explain reversibility, side effects, backup requirements and the smallest safe next action. Do not interpret this template as authorization to modify files, launch apps, use shell commands, delete anything or send an email. Seek explicit approval through the supported ChatGPT and local EXZA controls. STEP 5 — Verify and report If authorization is granted, execute only the approved operation and independently inspect the outcome using an available read-only tool. If something fails, stops being accessible or is unsupported, stop and explain the real error instead of claiming success. FINAL RESPONSE FORMAT Connection and scope; Evidence collected; Findings; Safe plan and approval request; Verified outcome; Unresolved questions. Keep the answer specific to this task and redact sensitive data.

EXZA Remote — guided, permission-aware workflow My objective: Group visible names in Downloads into folders and propose a before-and-after tree; do not move or delete any content yet. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Read [PROJECT] README and check which documented runtime versions can be verified safely on Linux. Never install missing packages automatically. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Draft full content for a text-only PDF about [TOPIC], verify allowed output path and check that the target is absent before approval. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Search [PROJECT] source for TODO, FIXME and BUG comments and group findings by module and potential impact. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Propose read-only PowerShell or cmd version checks for [TOOLS], verify shell permission first and never install anything. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Design a source, logs, exports and archive layout for [PROJECT] inside an allowed Windows directory. Check naming conflicts first. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare [DEVICE A] and [DEVICE B] only using genuinely available EXZA device information; select devices one at a time if needed. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare approved [DOC A] and [DOC B] within size limits and list added, removed and changed instructions. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Find configuration filenames inside the approved Windows project without printing credential files or personal secrets. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Diagnose the existing failed build output for [PROJECT] on Windows and propose the smallest authorized next test, not an automatic reinstall. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inspect only [APP LOG FOLDER] under an approved Linux workspace using safe listings and metadata. Do not access protected system directories. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inventory approved [PROJECT] components and propose migration to [DESTINATION] including permission, collision and rollback checks. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Locate [SERVICE PROJECT] within an authorized Linux workspace; do not expand the search to protected directories. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inspect space use inside authorized [FOLDER] on Windows by safe metadata or approved commands; do not clear caches automatically. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Check source and destination metadata for copying [PROJECT] to [DESTINATION] on macOS; account for tool size limits, symlinks and no-overwrite rules. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare paginated EXZA search results for [TERM A] and [TERM B], distinguishing overlap, unique matches and coverage limits. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inspect branches, working tree and uncommitted changes of authorized Windows repository [PROJECT] without checkout, fetch, merge or reset. Target device: my authorized Windows computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare visible filenames in [EXPORT FOLDER] with [EXPECTED REPORTS] without opening the report contents. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare [PROJECT] runtime requirements with actual approved version checks on Linux; do not use sudo or install packages. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Plan to copy [SOURCE FILE] to [DESTINATION FILE] within one authorized area after checking size and target collisions. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Read bounded chunks from authorized [LOG FILE] on Linux, redact secrets and group repeated error messages. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Replace exactly [OLD TEXT] with [NEW TEXT] in [FILE] only after verifying the file checksum and one unique match. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Inspect approved status/log evidence for user-owned [SERVICE]; do not restart a service or change systemd without approval. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Identify approved files in [FOLDER] older than [DATE] using metadata only; age does not authorize deletion. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Propose approved read-only Git status and diff checks in [PROJECT] on Linux; avoid exposing credential-bearing content. Target device: my authorized Linux computer. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Windows/Linux support varies by tool and native verification is still ongoing; do not claim unavailable desktop or Mail functionality. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Find [FILE] in the EXZA recycle bin, verify restore ID and destination, and guard against overwriting an existing item. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Analyze the last approved test job for macOS [PROJECT], including exit status and actual failing assertions. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Review [PROJECT] README and package scripts and propose exact build steps with file-write effects and approval checkpoints. Target device: my authorized Mac. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. For macOS-specific applications, confirm the required desktop or Mail capability is available and authorized. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Compare bounded extracts from [LOG A] and [LOG B], separating new failures, warnings and unchanged results. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

EXZA Remote — guided, permission-aware workflow My objective: Propose an exact command for [TEST] in [PROJECT], disclose side effects and start the EXZA job only after approval. Target device: the currently paired Windows, macOS or Linux device, only where EXZA tools support it. 1. Connection and capabilities. Check the paired-device status and operating system. List the specific EXZA tools you can actually use for this task, including read-only versus write/terminal/desktop capabilities. Check actual operating system and available EXZA tools; do not assume that a Mac-only GUI or Mail action works on other systems. If a required capability is not exposed, stop that part and explain the limitation. 2. Scope and evidence. Ask which exact project, file, folder, window or recipient I intend if a placeholder remains ambiguous. Start with the narrowest read-only inspection supported by the authorized scope. Respect search-depth, size and pagination limits. Never probe hidden folders, credentials or unrelated personal data. 3. Plan before action. Summarize the observed facts separately from hypotheses. Show the exact paths, commands, text changes, recipients or windows that would be affected, plus risks, reversibility and prerequisites. Do not treat this prompt as permission for a write, shell job, GUI interaction, deletion or outbound email. 4. Approval and execution. Ask for explicit consent before every sensitive operation and respect the native EXZA/ChatGPT approval flow. Only after approval, use the smallest supported operation. Never escalate privileges, bypass a refused tool or automatically retry destructive steps. 5. Independent verification. Use a suitable read-only EXZA check to confirm what actually happened. When data is unavailable, a command fails, or the platform does not support the operation, say so instead of inventing success. Return your answer under: Device and permissions; Observations; Proposed steps; Approval needed; Verified results; Open questions. Keep sensitive values masked and make each recommendation actionable.

Use EXZA Remote to check whether my paired computer is online. Show the connection status and do not change anything.

Show the operating system, CPU and memory details of my connected computer. Keep the report concise and read-only.

Explain which EXZA Remote capabilities you can use in this session and which actions require additional permission. Do not assume access that has not been granted.

Use EXZA Remote to identify the authorized workspace and list its immediate contents without opening private documents.

List the running processes on my connected computer, including process names and memory use. Do not close anything.

Read system information and the available running-process summary. Highlight notable memory use without claiming a complete performance diagnosis.

Check my EXZA Remote connection state, identify what is actually known about a disconnected device, and suggest safe steps to reconnect. Do not bypass pairing.

Show the redacted recent EXZA tool-call history if it is available. Summarize actions without revealing credentials or sensitive content.

Read the available local EXZA usage count and explain what it represents. Do not guess my subscription or monthly quota from this count.

List the visible files and folders on my Desktop through EXZA Remote. Do not open, move or delete anything.

Show the visible items in my Downloads folder, grouped into files and folders. Do not inspect their contents.

Get the size, type and modification time of [FILE NAME] in my authorized Documents folder. Do not open its contents.

Search my authorized Desktop for files whose names contain [KEYWORD]. Return the matching paths without opening the files.

Open [FILE NAME] in my authorized folder and summarize its text. Keep the file unchanged and respect the tool size limit.

Read [FILE A] and [FILE B] in the authorized workspace and show the differences in a clear comparison. Make no changes.

Create a folder named [PROJECT NAME] inside my authorized workspace after confirming the destination. Do not overwrite existing items.

Copy [SOURCE FILE] to [NEW FILE] within the allowed area. First check that the destination does not exist, then ask before copying.

List Downloads and propose a folder structure based only on file names. Show the plan first; do not move any files until I approve.

Rename [OLD NAME] to [NEW NAME] in the same authorized directory. Verify both paths and ask before making the change.

Review [TEST FILE] in my authorized area and ask for approval before moving exactly that item to the recoverable EXZA recycle bin. Never permanently delete it.

List the items in the EXZA recycle bin without restoring or deleting anything. Explain which item I would need to select for a restore.

Find [TEST FILE] in the EXZA recycle bin. Verify its restore destination and ask before restoring it, without overwriting any existing file.

Read [LOG FILE] from my authorized workspace in bounded chunks, starting at offset 0. Summarize errors and stop at the supported limits; do not read unrelated private files.

If permitted, read a small bounded base64 chunk of [BINARY FILE] and explain its header or structure. Do not attempt to decode unrelated secrets.

List [FOLDER] and inspect the timestamps of the visible files that I select. Return a chronological report without touching the files.

Search the authorized workspace for a folder called [FOLDER NAME]. Return matching paths and do not open its contents.

Find filenames containing [KEYWORD] in the authorized project. Group the matching results by directory.

Search the authorized project source files for [EXACT PHRASE]. Show matching paths and short context; avoid outputting secrets or credentials.

Search the allowed project for TODO and FIXME comments. Return the file paths and summarize the tasks without changing code.

Locate files with [CONFIG KEYWORD] in their names inside the authorized project, then ask which one I want to inspect.

Search the authorized source folder for the definition of [FUNCTION NAME]. Return candidate filenames and the relevant source section.

List the authorized project tree within tool limits and identify repeated filenames in different folders. Do not assume files with the same name have identical contents.

Read the available EXZA Commander search settings and explain their result-count and search-depth limits. Do not modify them.

Explain the current EXZA Commander search depth and suggest a safe value for [TASK]. Ask before changing the allowed preference.

Search for [KEYWORD] in the authorized project and review the available result pages until all accessible matches are summarized. Do not open unrelated files.

Use a read-only terminal command appropriate to my connected operating system to show the current working directory. Explain the output.

If my authorized folder is a Git repository, run a non-modifying Git status check and summarize changed and untracked files. Do not commit or reset.

Review the unstaged Git diff in [PROJECT PATH] using a read-only command. Summarize changes without printing secrets or modifying the repository.

Inspect the package manifest in my authorized Node.js project, then propose non-destructive dependency checks. Ask before installing or updating anything.

Show me the exact command you would run to test [PROJECT] inside the authorized workspace. Wait for my approval before executing it.

Check versions of the selected development tools [TOOLS] using harmless version commands supported on this computer. Do not install anything.

Read the error output from my approved build job [JOB ID] and explain the most likely cause. Do not apply fixes before I approve a plan.

Propose the exact safe test command for [PROJECT] and its expected effects. Once approved, start a monitored EXZA terminal job and return its job ID.

Check the status of EXZA terminal job [JOB ID]. Show current output and whether the job is still running, without sending new input.

List the EXZA-owned interactive process sessions and identify the one I name. Do not send input or stop any session.

Identify my EXZA-owned test job [JOB ID], describe what stopping it may interrupt, and request approval before stopping that specific job.

Before running [COMMAND] with EXZA Remote, explain its purpose, expected files affected and security risk. Do not execute until I explicitly approve.

Use EXZA Remote to list the open application windows on my Mac. Do not focus, close or type into any window.

List installed macOS applications available to EXZA Remote. Group them by name without launching anything.

Open the Notes application on my authorized Mac after approval. Do not create or edit a note.

Bring [APPLICATION NAME] to the foreground on my Mac. Do not interact with its documents or dialogs.

Open https://example.com in the default browser on my connected Mac after I confirm. Do not log in or submit forms.

Show the exact text you would type into [ACTIVE APP] on my Mac. Ask me to confirm the focused window and the text before typing; do not press Return.

Tell me what keyboard shortcut [SHORTCUT] may do in the currently focused Mac app. Ask before sending it through EXZA Remote.

Explain that the Mac clipboard may contain sensitive information. Ask for explicit permission before reading its contents with EXZA Remote.

Put this exact non-sensitive text into my Mac clipboard after I approve replacing the current contents: [TEXT].

Show which Mac applications or visible windows are running and distinguish windows from background processes. Do not terminate anything.

Read [NOTES FILE] from my authorized Documents folder and extract decisions, deadlines and action items. Do not change the source.

Read the README in my authorized project and produce a beginner-friendly explanation of setup, commands and open questions.

Based on files I explicitly authorize, draft a project checklist in this conversation. Do not write a new file unless I approve the destination and content.

Prepare a short report on [TOPIC] and show the full text first. Ask before saving it as [REPORT FILE] in my authorized workspace.

Draft the content for a plain-text PDF report titled [TITLE]. Show the text and destination first, then request approval to create it with EXZA Remote.

Read [FILE] in the authorized workspace and identify the single exact passage [OLD TEXT]. Show the replacement [NEW TEXT] and ask before applying a verified edit.

Check [LOG FILE] in my authorized workspace, show the line [ENTRY] to append, and request confirmation before adding it with version verification.

Read the explicitly selected text files [FILE A], [FILE B] and [FILE C] in the same authorized area, within the combined size limit. Summarize shared themes.

If [CSV FILE] is readable as text through EXZA Remote, inspect only an approved small portion and describe its columns. Do not modify the data.

Read the authorized source file [FILE], explain how to implement [FEATURE], and show the proposed change before editing anything.

Prepare an email preview to [ADDRESS] about [TOPIC]. Show the subject and full body. Do not create a draft or send anything yet.

Write a polite email draft to [ADDRESS] about [TOPIC]. Show it first and ask before opening the draft in macOS Mail.

Prepare an email to [ADDRESS] with the subject [SUBJECT] and body [MESSAGE]. Verify the recipient and show the final text. Send only after my explicit approval.

Rewrite this email text to be professional and concise: [MESSAGE]. Do not open Mail or send anything.

Draft a clear support reply explaining [ISSUE] and [SOLUTION]. Present the full preview without sending it.

Explain the difference between read-only access, file changes, terminal execution and desktop control in EXZA Remote. Tell me when local confirmation is required.

Tell me which folders and computer actions are authorized in this EXZA session. Do not try to bypass any blocked operation.

Run a safe connection check and a read-only file-listing test in the authorized workspace. Report actual success or errors without changing settings.

If available, summarize the redacted recent EXZA actions and explain which ones were read-only versus state-changing. Do not reveal private data.

Review only names in [FOLDER] and propose a cleanup plan. Never permanently delete anything; request approval for each move to the recycle bin.

Explain this EXZA permission error: [ERROR TEXT]. Suggest the supported authorization path. Do not propose ways to circumvent the restriction.

After an explicitly approved action, use an appropriate read-only EXZA tool to verify its result. Show the actual response, not an assumed success.

Before changing [TARGET], describe exactly which EXZA tool would be used, the required permission and the possible side effects. Wait for approval.

Compare which EXZA Remote functions are listed as Mac-specific and which may apply to Windows or Linux. Clearly mark anything not verified on my device.

Use EXZA Remote on my authorized Windows computer. Task: Check the authorized Windows PC for OS, architecture, CPU, RAM and uptime using the available read-only EXZA system information. Keep unknown values unknown. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Identify kernel, distribution if available, CPU, RAM and uptime for the authorized Linux device. Do not infer a distribution from the kernel alone. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Summarize the current Mac connection, uptime, memory and bounded process information. Distinguish observations from a performance diagnosis. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Find high-memory processes using the bounded EXZA process list on my Windows machine. Do not terminate or label them malicious. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Identify currently visible workloads and resource use on my Linux host without reading private process arguments or terminating anything. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Check the documented prerequisites for [PROJECT] on Windows; propose version checks for Node, Git and Python only where they are relevant and authorized. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Read [PROJECT] README and check which documented runtime versions can be verified safely on Linux. Never install missing packages automatically. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare [DEVICE A] and [DEVICE B] only using genuinely available EXZA device information; select devices one at a time if needed. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: List [PROJECT FOLDER], find README and entry-point filenames, then read only chosen small text files within EXZA limits. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Group visible names in Downloads into folders and propose a before-and-after tree; do not move or delete any content yet. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Design a source, logs, exports and archive layout for [PROJECT] inside an allowed Windows directory. Check naming conflicts first. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Inspect only [APP LOG FOLDER] under an approved Linux workspace using safe listings and metadata. Do not access protected system directories. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Check source and destination metadata for copying [PROJECT] to [DESTINATION] on macOS; account for tool size limits, symlinks and no-overwrite rules. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare visible filenames in [EXPORT FOLDER] with [EXPECTED REPORTS] without opening the report contents. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Plan to copy [SOURCE FILE] to [DESTINATION FILE] within one authorized area after checking size and target collisions. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Replace exactly [OLD TEXT] with [NEW TEXT] in [FILE] only after verifying the file checksum and one unique match. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Identify approved files in [FOLDER] older than [DATE] using metadata only; age does not authorize deletion. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Find [FILE] in the EXZA recycle bin, verify restore ID and destination, and guard against overwriting an existing item. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Find accessible README-like filenames under [PROJECT] using bounded EXZA search; show the paths grouped by subfolder. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Search allowed source files for exact [OLD VALUE] and identify candidates for manual migration, without changing every match blindly. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare key sets of the approved German and English localization files and identify missing keys without inventing translations. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Search [PROJECT] source for TODO, FIXME and BUG comments and group findings by module and potential impact. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Find configuration filenames inside the approved Windows project without printing credential files or personal secrets. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Locate [SERVICE PROJECT] within an authorized Linux workspace; do not expand the search to protected directories. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare paginated EXZA search results for [TERM A] and [TERM B], distinguishing overlap, unique matches and coverage limits. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Propose read-only PowerShell or cmd version checks for [TOOLS], verify shell permission first and never install anything. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Diagnose the existing failed build output for [PROJECT] on Windows and propose the smallest authorized next test, not an automatic reinstall. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Inspect space use inside authorized [FOLDER] on Windows by safe metadata or approved commands; do not clear caches automatically. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Windows computer. Task: Inspect branches, working tree and uncommitted changes of authorized Windows repository [PROJECT] without checkout, fetch, merge or reset. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Compare [PROJECT] runtime requirements with actual approved version checks on Linux; do not use sudo or install packages. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Read bounded chunks from authorized [LOG FILE] on Linux, redact secrets and group repeated error messages. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Inspect approved status/log evidence for user-owned [SERVICE]; do not restart a service or change systemd without approval. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Linux computer. Task: Propose approved read-only Git status and diff checks in [PROJECT] on Linux; avoid exposing credential-bearing content. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Analyze the last approved test job for macOS [PROJECT], including exit status and actual failing assertions. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Review [PROJECT] README and package scripts and propose exact build steps with file-write effects and approval checkpoints. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare bounded extracts from [LOG A] and [LOG B], separating new failures, warnings and unchanged results. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Propose an exact command for [TEST] in [PROJECT], disclose side effects and start the EXZA job only after approval. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: List visible application windows on the authorized Mac without focusing apps or reading their private content. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Check that [APP NAME] is installed, then request permission before launching it on the Mac. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Find the exact [APP NAME] window in the visible Mac window list and request approval before switching focus. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Prepare [TEXT] for the focused Mac app [APP] and request confirmation of target window and text before typing; do not press Return. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Prepare [TEXT] for the Mac clipboard and ask before replacing current contents; do not read the previous clipboard. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Explain the possible effect of [SHORTCUT] in focused [APP] and ask before sending any keyboard event. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Read approved [NOTES] and extract tasks, owners and due dates without inventing missing information. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Inspect only a bounded part of approved [CSV FILE], identify delimiter, columns and obvious missing values. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Turn [LOG] from an approved workspace into evidence-based expected/actual behavior and reproduction steps, removing personal data. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Use the approved README and tests of [PROJECT] to map passed evidence, missing checks and rollback steps without deploying. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Read the approved source section for [FEATURE], prepare a minimal diff and validation plan, and do not edit until approved. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Draft full content for a text-only PDF about [TOPIC], verify allowed output path and check that the target is absent before approval. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Compare approved [DOC A] and [DOC B] within size limits and list added, removed and changed instructions. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Inventory approved [PROJECT] components and propose migration to [DESTINATION] including permission, collision and rollback checks. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Draft a helpful reply to [QUESTION] with subject and body; never open Mail or send without separate approval. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Prepare a service notice for [AUDIENCE] using only confirmed scope and dates; show short and full variants without sending. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Review recipient [ADDRESS], subject [SUBJECT] and message [BODY] before irreversible sending on macOS Mail; require explicit approval. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on my authorized Mac. Task: Draft a follow-up on [REQUEST] based only on provided dates and facts; show the text before any Mail draft is created. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Explain the EXZA PIN and physical Y/N approval flow and which computer operations remain blocked without owner consent. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Assess [ACTION] for read, write, terminal and desktop permissions and show safer alternatives without executing it. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Check available connection state and explain local terminal stop, web disconnect and device removal without bypassing controls. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Review [PROMPT] for passwords, broad folder access and unnecessary data sharing, then rewrite it to minimize permissions. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

Use EXZA Remote on the currently paired Windows, macOS or Linux device, only where EXZA tools support it. Task: Verify approved [CHANGE] using read-only EXZA evidence; report success, failure or uncertainty without retrying writes automatically. First verify the connection, device permissions and supported tools. Work only within my authorized scope; show the plan before any state-changing or external action. Report the actual outcome, errors and anything you could not check.

EXZA Remote request for my authorized Windows computer: Help me prepare [WINDOWS PC] for an EXZA Remote connection without making system changes. Work plan: Check the actual available installer instructions, local owner confirmation and network reachability; distinguish unverified Windows functionality. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Create a step-by-step readiness checklist for linking [LINUX DEVICE] to EXZA Remote. Work plan: Use documented launcher and pairing flow, explain Y/N owner decision, do not assume root access or install packages. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Investigate why my paired Mac occasionally becomes disconnected without altering connectivity. Work plan: Collect EXZA status, recent authorized connection events and actual failure messages; distinguish pause, revoke, exit and network outage. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Compare the permissions and status of selected EXZA devices while protecting private device details. Work plan: Name the currently selected device, inspect only exposed scope and approval mode, never silently switch another owner account. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Create a diagnostic flow for [EXZA ERROR] with evidence-based decisions and no permission bypass. Work plan: Separate offline agent, denied scope, expired OAuth, unpaired device and unsupported action; propose only official recovery paths. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Inspect the names and sizes of [PROJECT BACKUPS] without unpacking or modifying archives. Work plan: Verify archive filenames, dates and apparent version labels in authorized folders; show naming inconsistencies. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Inspect metadata for [PROJECT FILES] inside the permitted Linux workspace. Work plan: Use EXZA metadata and supported read-only shell inspection only if authorized; do not run chmod or change ownership. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Inventory screenshot filenames in my authorized Pictures folder without reading image content. Work plan: Use a bounded file listing and metadata; group by date and name patterns; no deletion or upload. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Propose consistent safe names for copies of [PROJECT] before the next release. Work plan: Inspect existing filenames and versions; avoid overwrites and distinguish backup copies from production sources. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Analyze [LARGE TEXT FILE] using bounded EXZA file chunks with a final coverage report. Work plan: Verify file version and checksum at each continuation; respect offsets and file-size caps, avoid secrets and claim only read sections. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Verify whether [FOLDER] contains each named file from [MANIFEST] without creating or deleting anything. Work plan: Normalize filenames carefully, compare exact names and note unreadable entries rather than guessing they are absent. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Plan renaming selected [FILES] to [NEW NAMING RULE] without performing the batch yet. Work plan: Check each original, target collision and ordering; request item-level approval and never overwrite existing files. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Find where the expected [ARTIFACT NAME] should be inside the approved Linux project. Work plan: Compare expected build output locations from docs to bounded filename search; do not rebuild or generate files automatically. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Compare authorized [SOURCE FOLDER] with [DESTINATION FOLDER] by visible metadata and plan a collision-free reconciliation. Work plan: Distinguish missing, changed-size and same-name items; do not claim content equality without hashes and do not copy until approved. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Determine why [TEXT EXPORT] looks truncated or contains unexpected characters without modifying it. Work plan: Check size, encoding markers, accessible text chunks and observed errors; propose safe recovery only after evidence. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Locate the exact user-visible error [MESSAGE] in approved project sources and identify the surrounding feature. Work plan: Search literal text first, then relevant localization keys within bounded directories; report references and do not edit. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Find references to [IMAGE FILE] in authorized HTML/CSS/JS without opening private assets. Work plan: Search filename and related paths, group direct links from CSS and HTML and note dynamic references as uncertain. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Find which approved project files refer to its Windows installer [INSTALLER NAME]. Work plan: Inspect read-only filenames and safe text references; never execute installers or print embedded credentials. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Locate the instructions and scripts for [LINUX LAUNCHER] in an authorized repository. Work plan: Search known script names then README references; distinguish documented actions from tested operation. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Find files whose names suggest old [FEATURE] versions, without automatically classifying them as unused. Work plan: Check references within accessible source and mark uncertainty when search limits or dynamic imports prevent a complete answer. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Find the approved notes and scripts mentioning [MIGRATION NAME] and construct a chronology. Work plan: Search names and terms, inspect only selected small text files, sort by recorded dates rather than assumed chronology. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Identify accessible tests that mention [FUNCTION NAME] and distinguish references from actual executed coverage. Work plan: Search test filenames and text, relate matches to package scripts and report gaps without claiming tests passed. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Locate labels and aria attributes for [UI COMPONENT] within approved frontend source. Work plan: Search related component names in HTML and JS; report visible labels and missing obvious translations only from verified source. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Examine why [COMMAND] fails when a Windows project path contains spaces or special characters. Work plan: Inspect existing error, distinguish PowerShell from cmd quoting and propose harmless dry-run examples before any execution. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Investigate a permission-denied error in an authorized [PROJECT] build without raising privileges. Work plan: Read available error and path metadata, identify ownership hypothesis versus confirmed restrictions, avoid sudo, chmod or chown changes. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Inspect [PROJECT] requirements and safely verify the installed developer toolchain on the Mac. Work plan: Use approved version-only commands for tools named by the README, avoid installing Xcode components automatically. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Assess the exact proposed [ROLLBACK COMMAND] for [PROJECT] before anyone runs it. Work plan: Identify affected paths, destructive flags and existing backups; show a safer preview or dry-run if supported and await explicit approval. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Review npm scripts of [NODE PROJECT] on Windows and distinguish safe checks from commands that modify files. Work plan: Read package.json, classify each script, flag pre/post hooks and do not run install, build or clean without permission. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Review authorized Docker Compose or container definitions for [SERVICE] without launching containers. Work plan: Read safe source files only, identify ports, mounts and environment-variable names; mask values and avoid docker execution. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Triage failure output of the approved Windows [TEST RUN]. Work plan: Use existing bounded logs, separate failed assertions from skipped tests, propose a focused rerun only after approval. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Investigate why authorized [START SCRIPT] on macOS exits without showing the expected prompt. Work plan: Review shell syntax, permission metadata and actual captured output; avoid chmod and unattended relaunch until approved. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Compare two approved text configs [CONFIG A] and [CONFIG B] and summarize behavioral differences. Work plan: Mask credentials, compare key names and nonsecret values, do not apply new settings or restart any service. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Analyze the status and output of EXZA-owned job [JOB ID] without restarting or stopping it. Work plan: Use EXZA job status, timestamps and bounded output; distinguish running, failed and unknown results and list safe next actions. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Windows computer: Inspect [PYTHON PROJECT] for documented virtual-environment expectations on Windows. Work plan: Read requirements and metadata, propose safe Python version checks; creating a venv requires explicit approval. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Linux computer: Review an approved user-owned scheduled-task definition [SCHEDULE FILE] without installing or changing it. Work plan: Read only allowed text definition and interpret frequency, working directory and file effects; do not modify cron or timers. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Determine whether approved [APP NAME] can be launched by the available EXZA Mac desktop tools. Work plan: List installed apps and current windows, compare exact app name, then ask before initiating a launch. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Plan multiple text entries in [APP] without accidentally submitting or affecting a different window. Work plan: Confirm focus at each step, preview the exact text and shortcuts, request approval and stop on unexpected dialogs. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Identify which already installed Mac applications might be useful for [WORKFLOW], without launching them. Work plan: Use EXZA installed-app list, match names to the task and distinguish likely relevance from confirmed functionality. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Plan copying prepared [TEXT] and pasting into selected [APP] without reading unrelated clipboard data. Work plan: Confirm app window, show new text, explain clipboard overwrite and request separate write and paste permissions. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Summarize repeated issues in approved [ISSUE LOG] without assuming they are all still open. Work plan: Extract dates, observed failures, fixes and unverified status from bounded file sections; avoid exposing personal identifiers. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Draft a repeatable release runbook for [PROJECT] from approved documentation without executing deployment. Work plan: Include prerequisites, backup, dry-run, approval gates, verification, failure modes and rollback only where backed by source. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Create a beginner-friendly onboarding checklist for [PROJECT] based on its approved README. Work plan: Separate OS-specific commands, safety approvals and prerequisites; mark unknown steps and do not install software. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Compare permitted CSV snapshots [OLDER CSV] and [NEWER CSV] within file limits. Work plan: Verify common column names, unique row identifiers and data types; describe only the sampled portion if truncated. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Write a customer help article for [KNOWN ERROR] using evidence from approved error messages and documentation. Work plan: Give symptoms, affected platforms, safe checks, official recovery and escalation; do not recommend bypassing approval prompts. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Verify claims in [RELEASE NOTES] against approved source and test artifacts before publishing. Work plan: Match each claimed feature to documented code and test results; flag anything unverified instead of rewriting facts. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Prepare a Mac Mail follow-up to [CUSTOMER] about an incomplete EXZA setup without assuming the cause. Work plan: Ask for the exact observed error, avoid asking for passwords or PINs and provide safe troubleshooting steps; send only after confirmation. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for my authorized Mac: Prepare [RELEASE EMAIL] and verify new-feature claims against approved release documentation before sending. Work plan: Separate confirmed features, limitations and platform-specific notes; review final recipient and body and do not send automatically. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Assess risks of removing [BACKUP FILE] while keeping a recoverable option wherever EXZA permits. Work plan: Check file name, destination, other backup copies and explicit approval; prefer EXZA recoverable trash; never permanently delete silently. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Plan a least-privilege scope for EXZA access to [PROJECT] and its required tasks. Work plan: Identify read-only actions, exact permitted folders and separate approval needs for editing or running commands; do not alter security settings automatically. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote request for the selected authorized Windows, macOS or Linux machine when its EXZA tools support the operation: Evaluate an EXZA-assisted email workflow for accidental disclosure of private file content or addresses. Work plan: Check recipient, attachment scope and included details, prefer preview and redaction and demand final send confirmation. Start read-only within the exact approved folder or device, show any intended changes and ask for approval, then report only verified results. Do not expose secrets or work around rejected permissions.

EXZA Remote task — Dimensioned 3D cube. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check unit scale and specified dimensions; propose Blender Python to construct and verify bounding box. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Electronics enclosure. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Request PCB dimensions, wall thickness and screw spacing; make separate lid and base; check clearances. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Printable hinge. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Specify pin diameter, rotational clearance, print direction and manifold geometry; verify interference. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Servo bracket. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Require actual servo datasheet and measured mounting holes; model with test tolerance; export provisional STL. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Low-poly game props. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Define triangle budget, origin and asset sizes; create distinct meshes and material variants. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Atmospheric horror scene. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Block out terrain and props, set dramatic lighting and volumetrics, preview low-resolution composition. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Hard-surface bevels. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect applied scale and normals; add reversible bevel modifier; confirm target dimensions remain correct. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Brushed-metal material. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Build version-compatible shader nodes, roughness variation and directional texture; validate preview. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Studio product lighting. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Plan camera lens, background and key, fill and rim lights; test exposure and clean shadow. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Cycles render optimization. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check installed render engine and CPU/GPU support; tune samples, bounces and denoise with benchmark frames. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Eevee preview pipeline. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check Blender version, select supported Eevee engine, lower render scale, preserve production settings. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Camera orbit animation. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Define fps, pivot and orbit length; generate keyframes and validate smooth interpolation and last frame. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Product turntable animation. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Choose object or camera rotation, fix pivot, keep lighting stable and verify loop continuity. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — GLB export for web. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Validate mesh scale, material compatibility, texture paths and triangle budget; inspect exported file. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Watertight STL export. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check manifold edges, normals, disconnected shells and millimeter scale; advise slicer test. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Scene collection cleanup. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Audit object and collection names, propose naming rules, preserve parents and animation references. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Repair mesh normals. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect face orientation, custom normals and smoothing; repair only in a copy and compare results. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Geometry Nodes scatter. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Confirm API version and density; expose random seed and scale; preview with low instance count. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Parametric drilled plate. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Set thickness, hole diameter, center distance and edge margin; check clearances and dimensions. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — UV unwrap and texture checker. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect seams, island margins and distortion; display checker texture before texture painting. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Game mesh decimation. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Measure triangles, set a target budget, compare silhouette and normals against original. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Set pivots and transforms. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect object hierarchy and animation dependencies before changing origins or applying scale. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Basic mechanical rig. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Define joint axes, rotation limits and hierarchy; test movements for collision and distortion. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Emboss 3D labels. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check font licensing, stroke width, extrusion depth and print readability; preview placement. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Mechanical assembly fit check. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Align parts using measured references; inspect collisions, screw access and clearance; do not assert physical fit. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Transparent PNG product render. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Configure RGBA and film transparency, check shadows and alpha using a low-resolution preview. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Batch rendering scene variants. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check Blender CLI and render engine, estimate output size and show the exact batch command before approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Fix missing external textures. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inventory referenced textures within allowed paths; propose relative path repairs in a duplicate scene. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Compare two BLEND versions. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Use approved Blender Python metadata inspection for objects, mesh counts, materials and dimensions. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Prepare Godot GLB assets. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Check axis conventions, scale, origins and materials; export a copy and suggest import validation. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Prepare Unreal asset export. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Confirm Unreal import requirements, centimeters/meter conversion, collision naming and supported format. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — 3D printer tolerance coupon. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Model marked holes, pegs and slots with controlled tolerance increments; check watertightness before printing. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Reusable material swatch library. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Define PBR roughness and color variants; assign to test objects and compare renders consistently. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Pack scene dependencies. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inventory relative texture paths and external files; plan packing or copying only after approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Create a modular wall kit. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Specify grid dimensions, snap interfaces and texture scale; validate pieces align on a sample grid. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Build a simple staircase. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Calculate riser and tread dimensions from total rise and step count; generate measured stair geometry. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Create cable routing channel. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Ask for cable diameter, bend radius, wall thickness and mounting route; validate print clearances. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Render a clay concept preview. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Use a neutral clay material, soft studio light and consistent camera to inspect silhouette without costly materials. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Create a photogrammetry cleanup plan. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect holes, noisy surfaces, scale reference and mesh density; propose non-destructive cleanup steps. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Blender Python script code review. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Review bpy script for destructive operators, version mismatches, path writes and missing output validation before running. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Business introduction email. On the authorized Mac, Prepare a concise subject and introduction for recipient and service; do not invent testimonials or send automatically. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Answer invoice question. On the authorized Mac, Verify provided invoice number, payment record and date; draft a courteous response without assuming anything unpaid. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Attachment-ready email review. On the authorized Mac, Check recipient and attachment file metadata; explain that EXZA Mail tools may not support adding attachments. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Polite payment reminder. On the authorized Mac, Use only verified due dates and received payments; avoid threats and request approval before any send. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Ask for bug reproduction steps. On the authorized Mac, Ask the tester for OS, app version, redacted error and repeatable steps; never ask for passwords or PIN. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Bilingual release announcement. On the authorized Mac, Check actual release facts, maintain equal meaning across English/German and preview both drafts before sending. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Meeting request with agenda. On the authorized Mac, Write a meeting request with explicit timezone, three agenda points and optional slots without pretending calendar access. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Supplier quotation request. On the authorized Mac, List exact part dimensions, quantity and delivery question; claim drawings attached only if verified. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Sensitive support-mail screening. On the authorized Mac, Review recipient, pasted message and possible credential disclosure; redact data and do not send. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Service cancellation email. On the authorized Mac, Use confirmed contract reference and cancellation date; request written confirmation; avoid false legal claims. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Approve final outgoing email. On the authorized Mac, Read back full address, subject and body, explain irreversible sending, ask for explicit approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Email draft without mailbox access. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Write copy-ready subject and body in ChatGPT; do not claim EXZA can send on Windows/Linux. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Short project status email. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Summarize verified work, blocker and next milestone in under 120 words, without sending. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Apology for a delayed response. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Acknowledge delay, give realistic next step and omit invented deadlines; draft only. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Summarize pasted email thread. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Use only approved pasted thread, extract open questions, redact personal details and propose a reply. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Confirm a customer appointment. On the authorized Mac, Draft a clear confirmation with verified date, time, timezone and meeting format; confirm recipient before sending. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Respond to refund inquiry. On the authorized Mac, Draft a factual answer using confirmed policy and transaction facts; do not promise refund before approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Request a missing document. On the authorized Mac, Ask for a specific missing document using privacy-conscious transmission guidance and no unnecessary data. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Check email wording for legal claims. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Flag unsupported legal terms and deadlines in supplied draft, propose cautious wording and suggest professional review. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Follow up after software demo. On the authorized Mac, Summarize agreed next steps and ask for feedback without inventing customer interest or commitments. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Repeatable test pipeline. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Plan exact test commands, side effects, dependencies and success criteria; ask before running. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Monitor an EXZA test job. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Inspect a real authorized job ID and bounded output; do not invent background notifications. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Safe batch filename change. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Map source and destination names, check conflicts and rollback; get item-level approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Monthly report folder planning. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Review existing directory names, propose dated structure and create only after approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.

EXZA Remote task — Validate and zip build outputs. On the currently paired, authorized macOS/Windows/Linux computer, subject to its actual supported EXZA tools, Verify expected files and hashes, show exact archive command and check output after approval. First verify connection, access scope, installed tools and relevant source paths. Show a concrete plan and any script or command before changes; request explicit owner approval. Do not overwrite original files or send anything automatically. Return only actual verified results and next steps.