Use this skill to design, execute, and automate prompt regression test matrices and LLM-as-a-judge evaluation harnesses. It covers golden dataset curation, semantic embedding drift measurement, factual consistency scoring, and CI/CD gate automation before deploying prompt or model updates.
