requesting-code-review
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of Git diff outputs and project plan files during the review process. Malicious content within the code being reviewed could attempt to influence the logic or reporting of the sub-agents.
- Ingestion points: The skill reads data from
git diff --stat,git diff, and user-provided plan files (Step 2 and Step 3). - Boundary markers: The instructions do not define clear delimiters or "ignore instructions" warnings to separate the ingested code context from the sub-agent instructions.
- Capability inventory: The skill dispatches instructions to other agent instances using the
Agenttool, providing a mechanism for instructions in the data to potentially reach subsequent processing steps. - Sanitization: There is no evidence of sanitization, filtering, or escaping of the Git output before it is interpolated into the prompts for the specification or quality sub-agents.
Audit Metadata