The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands such as git diff, git log, git add, and git commit to analyze code changes and manage the refactoring process within the local repository.
[DATA_EXPOSURE]: The agent reads project configuration files and documentation to determine code idioms and team conventions. It specifically instructs the agent not to touch sensitive files like .env, secrets, or database migrations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase content and documentation, which constitutes an indirect prompt injection surface.
Ingestion points: Project configuration files (e.g., package.json), documentation (CLAUDE.md, MEMORY.md), and all source files within the audit scope.
Boundary markers: Strong instructions emphasize behavior preservation and alignment with existing project idioms.
Capability inventory: The skill can delete and modify files, and perform git staging/committing operations.
Sanitization: All changes must pass a verification step (/gate) including linting and unit tests, ensuring that any malicious or accidental behavior changes are caught before finalization.

audit