test-blindspots
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (specifications, code, tests) to identify logic gaps. 1. Ingestion points: Input from the user request, specification files, and existing code/tests. 2. Boundary markers: Use of specific scope and isolated data. 3. Capability inventory: Execution of test probes using existing tooling. 4. Sanitization: Instructions to keep probes away from production data and use local fixtures. The potential risk is mitigated by the skill's focus on isolated testing environments.
- [DYNAMIC_EXECUTION]: The skill involves designing and executing small exploratory probes (code snippets) to verify behavior. This dynamic generation is restricted to the context of the user's testing suite and local development environment, adhering to safe script execution practices for testing tools.
Audit Metadata