ai-system-testing
Pass
Audited by Gen Agent Trust Hub on May 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The static analysis detections for 'Ignore all previous instructions' and 'DAN' are false positives. These phrases are used explicitly as educational examples within the 'AI Safety Testing' section to describe adversarial inputs that a user should test against, rather than instructions to override the agent's behavior.
- [EXTERNAL_DOWNLOADS]: The skill recommends several well-known and reputable evaluation and red-teaming tools. These include:
- NVIDIA's Garak
- Microsoft's PyRIT
- UK AI Security Institute's Inspect AI
- Community-standard tools like Promptfoo, DeepEval, Ragas, and TruLens.
- [COMMAND_EXECUTION]: Provides standard CLI usage examples for the mentioned testing tools (e.g.,
pip install deepevalandgarakcommands). These are documented for the user's information and do not trigger automatic or hidden execution.
Audit Metadata