pull-request-review-gate
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted pull request content, which presents a surface for indirect prompt injection where malicious instructions could be embedded in code or descriptions to influence the review verdict.\n- Ingestion points: Untrusted data is ingested during Phase 0 (Establishing the Exact Review Subject) and Phase 2 (Reconstructing the Change) as described in
SKILL.md.\n- Boundary markers: The skill includes strong instructions for independence, specifically directing the agent to treat PR descriptions, comments, and prior reviews as claims until corroborated and to read requirements before the author's conclusions.\n- Capability inventory: The skill produces advisory review recommendations (APPROVE, REQUEST CHANGES) and deterministic scores, which are intended to inform merge readiness decisions.\n- Sanitization: The mandated 9-phase workflow requires independent reconstruction of failure paths and strictly forbids letting author-requested scores, confidence levels, or deadline pressure influence the audit-derived score.
Audit Metadata