CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918440640446464 |
|---|---|
| author | Zhang, Hanrong Fan, Shicheng Zou, Henry Peng Chen, Yankai Wang, Zhenting Zhou, Jiayu Li, Chengze Huang, Wei-Chieh Yao, Yifei Zheng, Kening Liu, Xue Li, Xiaoxiao Yu, Philip S. |
| author_facet | Zhang, Hanrong Fan, Shicheng Zou, Henry Peng Chen, Yankai Wang, Zhenting Zhou, Jiayu Li, Chengze Huang, Wei-Chieh Yao, Yifei Zheng, Kening Liu, Xue Li, Xiaoxiao Yu, Philip S. |
| contents | Anthropic proposes the concept of skills for LLM agents to tackle multi-step professional tasks that simple tool invocations cannot address. A tool is a single, self-contained function, whereas a skill is a structured bundle of interdependent multi-file artifacts. Currently, skill generation is not only label-intensive due to manual authoring, but also may suffer from human--machine cognitive misalignment, which can lead to degraded agent performance, as evidenced by evaluations on SkillsBench. Therefore, we aim to enable agents to autonomously generate skills. However, existing self-evolving methods designed for tools cannot be directly applied to skills due to their increased complexity. To address these issues, we propose CoEvoSkills, a self-evolving skills framework that enables agents to autonomously construct complex, multi-file skill packages. Specifically, CoEvoSkills couples a Skill Generator that iteratively refines skills with a Surrogate Verifier that co-evolves to provide informative and actionable feedback without access to ground-truth test content. On SkillsBench, CoEvoSkills achieves the highest pass rate among five baselines on both Claude Code and Codex, and also exhibits strong generalization capabilities to six additional LLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_01687 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification Zhang, Hanrong Fan, Shicheng Zou, Henry Peng Chen, Yankai Wang, Zhenting Zhou, Jiayu Li, Chengze Huang, Wei-Chieh Yao, Yifei Zheng, Kening Liu, Xue Li, Xiaoxiao Yu, Philip S. Artificial Intelligence Anthropic proposes the concept of skills for LLM agents to tackle multi-step professional tasks that simple tool invocations cannot address. A tool is a single, self-contained function, whereas a skill is a structured bundle of interdependent multi-file artifacts. Currently, skill generation is not only label-intensive due to manual authoring, but also may suffer from human--machine cognitive misalignment, which can lead to degraded agent performance, as evidenced by evaluations on SkillsBench. Therefore, we aim to enable agents to autonomously generate skills. However, existing self-evolving methods designed for tools cannot be directly applied to skills due to their increased complexity. To address these issues, we propose CoEvoSkills, a self-evolving skills framework that enables agents to autonomously construct complex, multi-file skill packages. Specifically, CoEvoSkills couples a Skill Generator that iteratively refines skills with a Surrogate Verifier that co-evolves to provide informative and actionable feedback without access to ground-truth test content. On SkillsBench, CoEvoSkills achieves the highest pass rate among five baselines on both Claude Code and Codex, and also exhibits strong generalization capabilities to six additional LLMs. |
| title | CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2604.01687 |