The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes git commands and the GitHub CLI (gh) to analyze code changes and fetch pull request data. These operations are performed as part of its primary function to review code and are restricted to the local working tree and official GitHub APIs.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically comments and reviews from GitHub PRs. Evidence: The skill uses gh api to harvest reviews in Step 2 and in the automated-reviewers.md reference. Capabilities: The skill can execute shell commands and modify local files like CLAUDE.md. Boundary markers: It implements instructions to distinguish bots by user.type and adheres to provenance rules to treat comments as data rather than instructions. Sanitization: It requires findings to be cross-referenced against the build plan and ledger before action is taken.

review-pass