ai-system-testing
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, credentials, or unauthorized network operations were detected. The skill provides a structured framework for testing AI features like LLM prompts and RAG systems.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to evaluate untrusted data from a System Under Test (SUT). Ingestion points: Untrusted data enters the context via SUT outputs, prompt configurations, agent tool call logs, and retrieval corpora (SKILL.md). Boundary markers: The methodology utilizes a strict rubric system and a judge calibration protocol to verify outputs (references/eval-design-patterns.md, references/judge-calibration.md). Capability inventory: The skill's primary capability is generating markdown reports; it does not perform subprocess execution, dynamic code evaluation, or direct network operations. Sanitization: Validation is enforced through structural assertions, key-field exact matches, and threshold-based scoring to sanitize the impact of non-deterministic or potentially malicious SUT outputs on the report.
Audit Metadata