flow-review
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data in the form of code changes, PRs, and repository metadata. This creates an attack surface where instructions embedded in code comments or diffs could attempt to influence the agent's behavior.
- Ingestion points: Phase 1 involves reading change requests, commits, and repository metadata.
- Boundary markers: The skill does not define specific syntax delimiters for the code under review.
- Capability inventory: The skill is purely instructional and does not contain subprocess calls, network operations, or file-write capabilities.
- Sanitization: The skill mitigates risks through its structured validation process in Phase 3, requiring findings to be refuted or confirmed through caller contracts, guards, and reachability analysis rather than accepting inputs at face value.
Audit Metadata