SkillsBench

SkillsBench is a benchmark dataset used to evaluate the performance of large language model (LLM) agents in executing tool-using tasks and assessing their ability to monitor and predict failures based on trace data.

Papers using SkillsBench (5)

MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation2026SkillRevise: Improving LLM-Authored Agent Skills via Trace-Conditioned Skill Revision2026SkVM: Compiling Skills for Efficient Execution Everywhere2026PrefixGuard: From LLM-Agent Traces to Online Failure-Warning Monitors2026SkCC: Portable and Secure Skill Compilation for Cross-Framework LLM Agents2026