ai-doc
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files to generate documentation, which constitutes an indirect prompt injection surface. Ingestion points: Accesses project files, source code, and commit history across the repository via tools like git and file walking. Boundary markers: Instructions explicitly separate read data from executable instructions and establish authority hierarchies for project context files. Capability inventory: Operations are limited to file system status checks and git-based file movement within a defined documentation scope. Sanitization: Directs the agent to detect, ignore, and report embedded imperatives found in content rather than acting on them.
Audit Metadata