llm-testing-expert

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely instructional, providing guidance on LLM testing methodologies like A/B testing, regression testing, and hallucination detection.
  • [NO_CODE]: No scripts, configuration files, or shell commands are included in the skill. It operates solely through text-based interaction schemas and workflows.
  • [PROMPT_INJECTION]: The content addresses security topics such as 'red teaming' and 'prompt injection' as subjects for testing and evaluation (defensive use case) rather than attempting to override the agent's internal safety protocols.
  • [DATA_EXFILTRATION]: There are no network operations, API calls, or file system access patterns present in the skill definition.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 11:48 AM
Security Audit — agent-trust-hub — llm-testing-expert