task-with-tests
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions focus on procedural tasks such as planning, writing tests, and building code based on user-accepted rules. All operations are local to the environment, using standard tools like git for tracking. No indicators of credential theft, data exfiltration, or persistence mechanisms were found.- [INDIRECT_PROMPT_INJECTION]: The skill consumes user input to determine which tests to implement, presenting a surface for indirect injection. Ingestion points: User replies to test selection questions in reference.md. Boundary markers: Decisions are isolated via a structured prompt batching system that requires explicit 'a' or 'b' responses. Capability inventory: Writing test and product code, and executing git add and git diff for file status. Sanitization: Human-in-the-loop gating ensures that only user-verified test logic is implemented.
Audit Metadata