review-the-diff
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from uncommitted code changes using git commands.
- Ingestion points: Code diffs retrieved via
git diffandgit diff --stagedas described inSKILL.md. - Boundary markers: No specific delimiters or instructions are provided to separate the diff content from the agent's instructions.
- Capability inventory: The skill's functionality is limited to analytical review; it does not possess tools for network access, arbitrary command execution, or file-system modifications.
- Sanitization: There is no evidence of sanitization or filtering for the ingested code content.
Audit Metadata