Agent Skills Evaluation Framework

Agent Skill evaluation harness for paired variants, trace artifacts, and runner adapters

Multi-tier framework for evaluating AI agent skills with quality gates, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior.