SelfEvoSkill revises agent skills via paired execution audits without external outcome supervision, reaching 73.9% mean reward across 89 tasks
The work introduces SelfEvoSkill, which pairs executions with and without a skill while holding agent, task, and workspace fixed, links the resulting behavioral differences to relevant skill passages, and iteratively revises and reevaluates those passages, with checks derived once from the task's explicit requirements preventing constraint violations; across 89 containerized tasks in 8 professional domains under the same best-of-two evaluation budget, it attains 73.9% mean task reward, exceeding the no-skill condition by +33.0 percentage points and the benchmark's curated skill by +17.2 percentage points, while skill evolution receives neither task rewards nor benchmark verifier outputs.