AI RESEARCH
SkillRevise: Improving LLM-Authored Agent Skills via Trace-Conditioned Skill Revision
arXiv CS.AI
•
ArXi:2606.01139v1 Announce Type: new Agent skills are procedural artifacts that enable LLM agents to execute workflows, verify constraints, and recover from failures. Existing self-evolving methods refine skills using accumulated trajectories. However, they struggle in cold-start settings, where only an initial, imperfect skill is available. Consequently, skill construction defaults to expert authoring or one-shot LLM generation.