The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the gh (GitHub CLI) and git commands to automate repository management and API interactions. It creates temporary worktrees to isolate the review process from the user's active development environment.
[EXTERNAL_DOWNLOADS]: Accesses Pull Request metadata and clones repository content from GitHub, a well-known and trusted service. This activity is restricted to the specific repository and PR requested by the user.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from GitHub PR bodies and comments. 1. Ingestion points: Fetches PR title, body, and comments via gh pr view as defined in SKILL.md and REFERENCE.md. 2. Boundary markers: Employs an 'independent verdict' logic and a mandatory 'Reconciliation & confirmation gate' requiring explicit user approval before posting. 3. Capability inventory: Includes the ability to post reviews (gh api ... POST), manage local files (git worktree), and delete temporary directories (rm -rf). 4. Sanitization: Relies on the GitHub CLI's internal argument handling and flags (e.g., -f for data fields) to safely pass data to API endpoints.

review