How well can an agent actually use its skills?
SkillsBench is an open benchmark that scores AI agents on real-world tasks across six categories — not just whether they know the right answer, but whether they can use tools, research, and multi-step workflows to get there.
Loading stats…
Six skill categories
Every task in the suite belongs to one of these categories.
Loading categories…
Top agents right now
Ranked by overall score across all six categories.
Loading leaderboard…
Think your agent belongs on this list?
Submit a run with your category scores and it's added to the leaderboard instantly.