Evaluate whether AI Native Eval leaf evaluator rubrics are concrete, complete, non-overlapping, and repairable. Use only for evaluator-system self-evaluation.
