The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts located in the .plan/ directory (e.g., execute-script.py). These are used for internal state management, such as computing git diffs, resolving architectural modules, and logging linting findings. These operations are restricted to the local worktree and the specific plan context.
[DATA_EXPOSURE]: The skill reads the 'footprint' (the files changed in a git diff) to perform a code review. This is the intended purpose of the skill. It does not attempt to access sensitive system files like credentials, SSH keys, or environment variables outside the project scope.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code files during the 'cognitive review' step. It includes instructions for the agent to use a specific 'reviewer lens' and strictly emit findings as a lint-issue type. While it processes external data (source code), the risk is mitigated by the structured output requirement (manage-findings add) and the workflow's narrow scope.
[REMOTE_CODE_EXECUTION]: No remote code execution was detected. All script executions target local files within the .plan directory, and there are no network-based downloads or curl|bash patterns.

recipe-code-review