review
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns or behaviors were found. The skill instructions adhere to high-assurance engineering principles, such as inspecting state before acting and reconciling contradictions between documentation and reality.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically code diffs and repository metadata, which creates a potential attack surface. The risk is mitigated by the skill's design and instructions.\n
- Ingestion points: Untrusted data enters via repository diffs and project specifications as described in SKILL.md and reality-check.md.\n
- Boundary markers: No explicit delimiters are defined to separate instruction context from external diff content.\n
- Capability inventory: The skill enforces a read-only-by-default policy and restricts modifications to those explicitly requested by the user, as defined in SKILL.md and diff-discipline.md.\n
- Sanitization: No explicit sanitization or filtering of external content is mentioned.
Audit Metadata