skill-creator
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data to generate new agent instructions.
- Ingestion points: The skill reads existing project code in Step 2 to detect patterns and extract examples for inclusion in new skills.
- Boundary markers: The skill lacks explicit instructions to ignore or sanitize embedded instructions within the code it analyzes.
- Capability inventory: The skill has the capability to write files to the global agent directory (
~/.claude/skills/) and modify project registry files (CLAUDE.md) in Step 5. - Sanitization: No escaping or filtering mechanisms are described for the content extracted from the project code before it is interpolated into the new skill skeletons.
- [PERSISTENCE]: The skill is designed to modify the agent's environment across sessions by creating persistent instruction files in
~/.claude/skills/and registering them in global configuration files. While this is the intended primary purpose, it allows for the permanent alteration of agent capabilities and behavior.
Audit Metadata