About shimo4228/llm-as-judge
An Agent Skill for anyone building an LLM-as-judge evaluator, such as a quality gate, a review step or a judge prompt. It fixes three chronic failures of such evaluators (scores that drift between runs, fatal defects diluted by averaging, and verdicts nobody can explain) with one design rule: collect evidence with binary Yes/No checks, let the LLM issue one named holistic verdict, and never aggregate the answers into a score.