ship
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing external specification files to drive high-capability sub-agents.
- Ingestion points: The skill accepts a path to a SPEC file (typically Markdown) which is then used to prompt sub-agents in the Prepare, Implement, and Verify phases (e.g., in
assets/workflow.js). - Boundary markers: The workflow script does not use explicit delimiters or "ignore instructions" wrappers when interpolating SPEC content into sub-agent prompts.
- Capability inventory: The agents have access to tools that can modify the repository (
implement-with-tdd), execute the project's code (runskill), and create GitHub pull requests (prskill). - Sanitization: There is no evidence of input validation or sanitization on the SPEC content before it is processed by the LLM agents.
Audit Metadata