agent-automation-recommender
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_CONTEXT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads codebase manifest files (e.g., package.json, Cargo.toml) and source directories to identify frameworks and libraries. This ingestion of external data creates a surface for indirect prompt injection, although the skill's output is limited to user recommendations.\n
- Ingestion points: Project manifest files (package.json, pyproject.toml, etc.) and directory structures.\n
- Boundary markers: Not explicitly present in file-reading commands.\n
- Capability inventory: Uses Read, Glob, Grep, and Bash for reconnaissance.\n
- Sanitization: No explicit filtering of file content before analysis.\n- [COMMAND_EXECUTION]: The skill uses common system commands (ls, cat, grep) to perform codebase reconnaissance and detect the project's environment stack.\n- [EXTERNAL_DOWNLOADS]: Reference guides provide instructions for integrating well-known services and official plugins (e.g., Supabase, Convex, GitHub), which are considered safe sources.\n- [DYNAMIC_CONTEXT_INJECTION]: The skill reference material documents the use of dynamic context injection (e.g.,
!command) as an example for users creating their own skills, which is handled as a legitimate educational use case.
Audit Metadata