maintain
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external sources (GitHub issues, PR descriptions, and comments) which could contain malicious instructions.
- Ingestion points: Fetches live state from GitHub including issue/PR descriptions, authors, and review comments in the 'Pick the next item' and 'Decide one item' sections.
- Boundary markers: Implements a 'Review boundary' containing five specific validation questions to verify if PRs match their declared contracts and problems.
- Capability inventory: Employs the 'gh' CLI to perform mutations such as labeling, commenting, closing issues, and squash-merging PRs. It also performs code changes in isolated worktrees.
- Sanitization: Relies on logical validation (Review boundary) rather than technical sanitization of the input text.
- [COMMAND_EXECUTION]: The skill utilizes the GitHub CLI ('gh') to interact with repository data and perform administrative actions.
- The instructions explicitly authorize the use of 'gh pr', 'gh issue', 'gh run', and 'gh api' for triaging the 'Q00/ouroboros' repository.
- It includes a safety check using 'gh auth status' to verify permissions before attempting mutations.
Audit Metadata