让下一篇重要论文，不再错过。

原文摘要

This paper proposes EvoAgent--an evolvable large language model (LLM) agent framework that integrates structured skill learning with a hierarchical sub-agent delegation mechanism. EvoAgent models skills as multi-file structured capability units equipped with triggering mechanisms and evolutionary metadata, and is designed to enable continuous skill generation and optimization through a user-feedback-driven closed-loop process. In addition, by incorporating a three-stage skill matching strategy and organizing the agent as a harness whose only self-modified component is a persistent, user-owned state (profile, memory and skills), the framework supports dynamic task decomposition for complex problems and long-term capability accumulation. Experimental results based on real-world foreign trade scenarios demonstrate that, after integrating EvoAgent, GPT-5.2 achieves substantial improvements in professionalism, accuracy, and practical utility. Under a five-dimensional LLM-as-Judge evaluation protocol, the overall average score increases by approximately 28%. Further model transfer experiments indicate that the performance of an agent system depends not only on the intrinsic capabilities of the underlying model, but also on the degree of synergy between the model and the agent architecture. Code and documentation are released at https://github.com/Focus-AI-Center/Mentarc-EvoAgent; the evaluation corpus contains proprietary business data that cannot be released, so the reported results are for reference only.