Enterprise work and professional deliverables
SkillsBench v1.1
87-task snapshot for testing the contribution of agent skills.
About SkillsBench v1.1
What it measures
Task reward; compare with and without the supplied skills.
Scope & limitations
A skills-effect experiment requires a paired baseline; an absolute score alone does not prove that skills helped.
Evaluation environment
Identical agent/model/runtime in paired skill conditions.