test-design-reviewer
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by reading test files and implementation code, which creates a potential attack surface for instructions hidden in code comments or data to influence the agent.
- Ingestion points:
SKILL.mdinstructs the agent to "Read the tests before the implementation" and "Inspect the public production boundary". - Boundary markers: No specific delimiters or warnings are provided to help the agent distinguish between its instructions and the content of the files being reviewed.
- Capability inventory: The skill allows the agent to "Run focused tests" using local development tools.
- Sanitization: There is no mention of sanitizing or escaping the content of the files before processing them.
- [SAFE]: The skill includes references to external websites (LinkedIn) and public code repositories (GitHub) for attribution and historical context. These links point to well-known services and the author's own files, representing legitimate documentation practices.
- [SAFE]: The review process explicitly requires the agent to "Run focused tests or timing only when authorized," ensuring human-in-the-loop control over execution.
Audit Metadata