setup-engine
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the web (migration guides, breaking changes) to update project documentation and agent instructions. \n
- Ingestion points: Content retrieved via
WebFetchfrom external documentation sites (Sections 7, 10, 11). \n - Boundary markers: None identified. The skill directly populates documentation files and modifies agent instructions based on external content. \n
- Capability inventory: The skill has
WriteandEditaccess toCLAUDE.md, specialist agent files, and project documentation directories. \n - Sanitization: No explicit sanitization or validation of the fetched external content is performed before use. \n- [EXTERNAL_DOWNLOADS]: Uses
WebSearchandWebFetchto gather engine-specific technical data, such as versions, migration guides, and breaking changes from external websites to supplement the agent's knowledge.
Audit Metadata