re-extract-requirements
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates within a defined workspace (
docs/reverse/) and uses standard file system tools for reading source code and writing markdown reports. It does not perform network operations, access sensitive credentials, or execute untrusted code. - [INDIRECT_PROMPT_INJECTION]: The skill processes external source code and logic diagrams, which could theoretically contain malicious instructions. However, the instructions explicitly mandate a verification step (Step 3: Critic) that requires the agent to re-read the source file and verify evidence for every claim. This robust verification process, combined with the language-agnostic extraction focus, effectively mitigates the risk of indirect prompt injection.
Audit Metadata