Skills
skills/latestaiagents/agent-skills/eval-dataset-design/Gen Agent Trust Hub

eval-dataset-design

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely educational, providing guidelines on dataset coverage, labeling quality, and contamination checks.
  • [REMOTE_CODE_EXECUTION]: Includes illustrative code examples using the standard scikit-learn library for statistical metrics and common SDK patterns for data generation. These are provided as documentation and do not involve execution of untrusted code or remote scripts.
  • [DATA_EXFILTRATION]: No network access or sensitive file access patterns were identified in the instructions or examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 10:27 AM
Security Audit — agent-trust-hub — eval-dataset-design