Validate AI agent outputs: schema conformance, factual consistency checks, hallucination scoring, completeness assessmen