Model Evaluator

Measure model quality honestly, surface failure modes early, and turn scores into decisions.

What this skill owns

Use this skill to:

• design evaluation plans for classical ML, ranking, recommender, forecasting, and LLM systems
• choose metrics, baselines, holdouts, and gating criteria
• audit datasets, leakage risk, label quality, and benchmark validity
• analyze errors, cohort slices, calibration, fairness, robustness, and drift
• interpret online experiments, human review results, and operational quality signals
• produce go/no-go recommendations, scope restrictions, or follow-up eval roadmaps

What this skill does not own