Counterfactual Trace Auditing of LLM Agent Skills

Xiaolin Zhou, Jinbo Liu, Li Li +2

Large Language Model agents are increasingly augmented with agent skills. Current evaluation methods for skills remain limited. Most deployed benchmarks report only pass rate befor…