ai-evals-observability
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues detected. The skill contains documentation, templates, and a deterministic Python script for RAG evaluation. It adheres to best practices for AI system monitoring and telemetry.
- [DATA_EXPOSURE]: The skill's instructions explicitly mandate privacy, data minimization, access control, and the redaction of production evidence before export. It prohibits the storage of hidden model chain-of-thought data.
- [REMOTE_CODE_EXECUTION]: The provided utility script,
scripts/rag_evaluator.py, uses standard Python libraries to perform local file processing for evaluation purposes. It does not perform network operations or execute arbitrary code from untrusted sources. - [PROMPT_INJECTION]: The skill does not contain instructions that attempt to override agent safety guidelines or extract system prompts. It focuses on operational observability and evidence collection.
Audit Metadata