Use when a human wants to design, evaluate, or improve AI agent skills for scientific, legal, medical, education, engineering, finance, or other expert-domain workflows that require domain constraints and careful validation
