audit-pr
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the GitHub CLI (gh) to retrieve PR data, post comments, and toggle draft status. It also executes local JavaScript files via the Bun runtime (bun scripts/audit-pr-gate.mjs, bun scripts/review-receipt.mjs) for validation logic.
- [INDIRECT_PROMPT_INJECTION]: The skill audits untrusted data from the forge (PR bodies, comments) and local project files to determine readiness.
- Ingestion points: PR metadata, comments, and planning artifacts (e.g., PLAN.md, TASKS.md) accessed via gh and file system reads.
- Boundary markers: Employs SHA-bound HTML markers () to ensure that review receipts match the current state of the code.
- Capability inventory: Capability is limited to PR status updates (gh pr ready), posting verdict comments (gh pr comment), and writing to local audit logs (review-findings.md).
- Sanitization: Explicitly instructs the agent to use temporary files and the --body-file flag when posting content to the forge to prevent shell command injection through unsanitized text.
Audit Metadata