llm-wiki-eval
Pass
Audited by Gen Agent Trust Hub on Aug 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is entirely instructional and does not contain any executable scripts, command-line operations, or automated tool invocations.\n- [SAFE]: The methodology documentation provides a proactive security framework, instructing the agent to specifically audit for prompt injection, PII leakage, and unauthorized data access.\n- [SAFE]: The skill establishes safe data boundaries by recommending human-in-the-loop calibration, explicit safety gates for high-risk content, and data redaction policies.\n- [SAFE]: All external resource references target trusted academic repositories (ArXiv), well-known technology organizations (Amazon Science, OpenAI), or established open-source evaluation tools (promptfoo, Ragas, LangSmith).
Audit Metadata