address-pr-feedback
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from GitHub PR comments and reviews fetched via
gh apiand GraphQL. It addresses this risk by instructing the agent to 'adversarially verify' comments and treat them as unproven claims rather than instructions. - Ingestion points: SKILL.md identifies queries to fetch comments and reviews using
gh api. - Boundary markers: Strong instructions to distrust external input and require refutation before accepting changes.
- Capability inventory: Uses
gitandghfor codebase modification and PR management. - Sanitization: Mandatory logic to restate claims as triggers/consequences and verify against the head codebase.
- [COMMAND_EXECUTION]: Employs
gitfor commits and rebasing, and theghCLI for PR operations and stack management. These tools are used within their standard context for PR maintenance.
Audit Metadata