📑 arXiv 1w ago
SkillClaw: Let Skills Evolve Collectively with Agentic Evolver
SkillClaw enables collective skill evolution across multi-user LLM agent ecosystems by continuously aggregating interaction trajectories and autonomously refining skills via an agentic evolver, achieving 88% improvement after 6 rounds and +42.1% on real-world tasks. It enables cross-user knowledge transfer without additional user effort, solving the inefficiency where users repeatedly develop similar workflows independently.