The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local Node.js script located at .claude/hooks/lib/review-receipt.cjs to manage commit snapshots and issue review receipts. This involves executing code residing within the user's project directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the project's source code and git diffs during the review process, creating a surface for injection.
Ingestion points: Reads project source files, changed files, and commit descriptors in SKILL.md and references/fix-loop.md.
Boundary markers: The instructions do not specify any delimiters or safety warnings to ensure the agent ignores embedded instructions within the code it reviews.
Capability inventory: When the --fix-loop flag is used, the agent has the capability to write to the file system to apply fixes and execute the local review-receipt.cjs script.
Sanitization: There is no mention of sanitizing or validating the ingested code content before it is processed by the agent.

commit-review