Agent Skills下游效用的实证研究
原标题：An Empirical Study of Agent Skills' Downstream Utility
AI 摘要
该研究对87个SkillsBench任务进行实证研究，将下游效用定义为同一模型-执行框架配置下使用Skill与不使用Skill的通过率差异。研究发现同一Skill在36.78%的任务上对部分配置有帮助、对另一些配置有害，推荐流程可能成为执行负担，且相关性排序会忽略更有用的候选Skill。按所需操作支持重排序可在三种配置下将首选通过率提升4.35-5.80个百分点，并总结出17条编写实践，指出Stage Plan和Dependency DAG优于仅按使用顺序组织。
正文节选
An Empirical Study of Agent Skills’ Downstream Utility Abstract. Agent Skills package procedural guidance and resources for reuse, but a relevant Skill does not necessarily improve task performance. Existing studies characterize Skill content and evaluate downstream performance, yet provide limited explanations of how utility depends on content, execution configuration, and multi-Skill organization. We conduct an empirical study on 87 SkillsBench tasks, defining downstream utility as the pass-ra