Do "Agent Skills" actually make your LLM agents perform better? Researchers from BenchFlow and a diverse team from multiple institutions present SkillBench, a rigorous benchmark of 86 tasks across 11 domains. It precisely measures how well 'Agent Skills'—structured procedural

→ View original post on X — @jiqizhixin,