Formal Skill: Programmable Runtime Skills for Efficient and Accurate LLM Agents
The paper introduces Formal Skill, a new abstraction for enhancing the efficiency and accuracy of Large Language Model (LLM) agents. This approach utilizes JSON metadata and action schemas to create reusable capabilities, improving control and execution. The implementation, called FairyClaw, demonstrates competitive performance with reduced token usage in various tasks.
- ▪Formal Skill represents reusable capability with JSON metadata and action schemas.
- ▪FairyClaw is an open-source event-driven runtime for executing Formal Skills.
- ▪The new abstraction allows for token-efficient and enforceable control surfaces.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.19604 (cs) [Submitted on 19 May 2026] Title:Formal Skill: Programmable Runtime Skills for Efficient and Accurate LLM Agents Authors:Xi Zhang, Meijun Gao, Yuntian Zhao, Xinyu Tan, Yilun Yao, Feiyu Wang, Yanshu Wang, Dingsiyi, Tong Yang View a PDF of the paper titled Formal Skill: Programmable Runtime Skills for Efficient and Accurate LLM Agents, by Xi Zhang and 8 other authors View PDF HTML (experimental) Abstract:Large Language Model (LLM) agents increasingly act inside real workspaces, where tools and skills determine whether model reasoning becomes reliable action.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
