SkillOpt-Lite：通过一行氛围指令实现更优更快的智能体自我进化

摘要

尽管面向自主代理的技能优化已逐渐兴起，现有方法仍依赖于复杂流水线。其中一个基本问题尚未解答：构成最小可行技能优化流水线的要素是什么？其中每个组件需有理论或实证必要性的支撑。我们通过零阶优化形式化技能优化，将经典对应方法（中心差分、信任域）映射至近期文献。值得注意的是，与经典零阶优化中盲目的数值扰动不同，技能轨迹可作为可解释的调试反馈。基于Claude Code哲学与PAC学习理论，我们确立了关于收敛性与泛化性的三项原则：基于文件系统的轨迹探索、共识属性挖掘以及独立验证门控。通过消除冗余，我们提出SkillOpt-Lite。该方法加速收敛且全面优于完整版SkillOpt：在GPT-5.5上使LiveMath提升+8.8分，在GPT-5.4-nano上提升+25.4分，使得nano模型超越经SkillOpt优化的标准GPT-5.4。最后，我们将该框架集成至VSCode Copilot等生产级编码代理中，使开发者能通过一行“vibe”代码进化代理技能。由于我们的框架将所有代理组件视为标准可编辑代码，这一最小流水线自然可泛化至全流程优化（HarnessOpt）。在SpreadsheetBench上，HarnessOpt使GPT-5.4-nano达到0.7758的准确率，超越运行标准流水线的更大模型GPT-5.5（0.7620）。代码已开源：https://github.com/EvolvingLMMs-Lab/SkillOpt-Lite。

English

While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question unaddressed: What constitutes a minimal viable pipeline for skill optimization, where every component is justified by theory or empirical necessity? We formalize skill optimization via Zeroth-Order (ZO) optimization, mapping classical counterparts (central difference, trust regions) to recent literature. Noting that unlike blind numerical perturbations in classical ZO, skill trajectories serve as interpretable debugging feedback. Grounded in Claude Code philosophy and PAC learning, we establish three principles for convergence and generalization: file-system-based trajectory exploration, consensus attribute mining, and independent validation gating. Eliminating redundancies, we propose SkillOpt-Lite. It accelerates convergence and outperforms full SkillOpt: improving LiveMath by +8.8 points on GPT-5.5 and +25.4 points on GPT-5.4-nano, allowing the nano model to surpass standard GPT-5.4 optimized by SkillOpt. Finally, we integrate our framework into production coding agents like VSCode Copilot, enabling developers to evolve agent skills via one line of vibe. Because our framework treats all agent components simply as standard editable code, this minimal pipeline naturally generalizes to full harness optimization (HarnessOpt). On SpreadsheetBench, HarnessOpt enables GPT-5.4-nano to achieve 0.7758 accuracy, outperforming the larger GPT-5.5 running standard pipelines (0.7620). Code is available at https://github.com/EvolvingLMMs-Lab/SkillOpt-Lite.