untangle-codebase
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [SAFE]: The skill provides legitimate instructions for code refactoring and structural analysis. It includes explicit safety boundaries such as working only within authorized repositories and prohibiting the upload of private source code to external services without permission.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the codebase it is tasked to refactor (ingestion point: codebase files). While it lacks explicit prompt-level boundary markers or sanitization for embedded instructions in the source code, it relies on structured workflows and subagent reviews to verify changes. The capability to write files is the primary intended function of the skill.
- [DYNAMIC_EXECUTION]: The instructions suggest that the agent generate scripts for tasks like stripping function bodies or mapping dependencies to fit the codebase into context. These are standard development practices for large-scale code analysis and represent low-risk script generation from known templates rather than executing untrusted remote code.
Audit Metadata