review-feedback

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, which constitutes a potential attack surface. However, it incorporates strong defensive logic to mitigate this risk.
  • Ingestion points: The skill processes external code review feedback, PR comments, and findings from other models or review tools (as described in the YAML frontmatter and Sections 1 and 6).
  • Boundary markers: The skill explicitly establishes logical boundaries by requiring the agent to 'Adjudicate the reported problem and the proposed remedy separately before editing' (Header) and to 'Map every original comment to a group so none becomes an implicit patch instruction' (Section 4).
  • Capability inventory: The skill utilizes file-writing capabilities to implement code changes (Section 6) and shell execution capabilities through the tdd tool and issue reproduction steps (Section 2 and 6).
  • Sanitization: The skill mandates a verification process ('Try to disprove the claim', Section 2) where the agent must open cited code, trace execution paths, and check counterevidence before accepting any external claim. It explicitly warns that 'A real symptom does not make the reviewer's explanation or patch correct' (Section 2).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 04:33 PM
Security Audit — agent-trust-hub — review-feedback