pr-review-lead
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external sources, which could contain malicious instructions intended to influence the agent's behavior.
- Ingestion points: The skill reads pull request metadata, code diffs, and external repository doctrine files using
gh pr view,gh pr diff, andgh apicommands. - Boundary markers: The instructions do not define specific boundary markers or "ignore" delimiters for the content being reviewed in the diffs.
- Capability inventory: The agent has the capability to write back to external repositories by posting comments via
gh pr comment. - Sanitization: The skill implements strong human-in-the-loop controls. Step 5 (Triage) and Step 7 (Confirm, then post) require the user to manually review, edit, and explicitly approve the feedback before it is transmitted to GitHub.
- [COMMAND_EXECUTION]: The skill makes extensive use of the GitHub CLI (
gh) to interact with repositories. While these are intended operations for a PR review tool, they involve executing shell commands that interface with external services and file systems based on user-provided repository paths.
Audit Metadata