reverse-engineering-specs
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides a logical workflow for code analysis without including external code, third-party dependencies, or suspicious commands. It focuses entirely on documentation and analysis within the agent's provided environment.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase content which represents a vulnerability surface. However, it incorporates strong sanitization practices as part of its primary function.
- Ingestion points: The entire project codebase, including all entry points and data flows analyzed in Phase 1.
- Boundary markers: Hard-gates for traceability and mandatory transformation rules in Phase 2.
- Capability inventory: Use of the platform's Agent tool for exploration and local file system writes for specification storage.
- Sanitization: Mandatory requirements to strip all code, technical names, and implementation specifics, effectively sanitizing the input before it is stored in output files.
Audit Metadata