review-hog-resolution-criteria
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from pull request review threads. While a malicious actor could embed instructions in comments, the skill mitigates this by requiring the agent to independently verify claims against the code and existing tests, and by mandating human escalation for any behavior not provable in-session.
- Ingestion points: Pull request conversation threads and review comments (SKILL.md).
- Boundary markers: Absent; the skill relies on the agent's ability to distinguish commentary from code instructions.
- Capability inventory: The agent has the ability to write fixes to the codebase and commit changes (SKILL.md).
- Sanitization: Absent; the skill relies on logic-based validation and human review checkpoints.
Audit Metadata