●builderThis helps you diagnose specific failure points in your agentic workflows rather than just measuring final output accuracy.
●researcherYou can use this taxonomy to move beyond single-turn QA and study how reasoning errors cascade in legal agents.