migration-planner
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation includes validation commands for developers to run using internal platform scripts located in
$CODEX_HOME. These are standard maintenance procedures and do not involve untrusted remote sources or runtime execution by the agent itself.\n- [DATA_EXPOSURE]: Safety rules explicitly forbid the storage of sensitive information such as analytics tokens, cookies, or verification codes in the resulting migration plans, adhering to security best practices for credential handling.\n- [PROMPT_INJECTION]: The skill incorporates explicit safety instructions and a 'Safety Contract' that prevent the agent from executing production mutations (DNS changes, redirects, etc.) without clear user authorization, effectively mitigating risks of unauthorized actions.\n- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external site audit data, it lacks the technical capabilities to execute code or modify production environments autonomously.\n - Ingestion points: Site audits, URL inventories, and metadata preservation notes processed during the migration planning phase.\n
- Boundary markers: Instructions mandate that all production-related actions are marked as 'Planned' and require explicit human approval before execution.\n
- Capability inventory: No tools for file system writes, network requests, or subprocess execution are defined for the agent's autonomous use within the skill definition.\n
- Sanitization: All output is intended for human review in a plan format before any application, providing a manual verification gate.
Audit Metadata