Demystifying Agent Skills: Why They Work-Until They Don't

  • 类型：arxiv
  • 标识：2608.14036
  • 链接：https://arxiv.org/abs/2608.14036
  • 主分类：agent
  • 形态：benchmark
  • 被引：0
  • 被引来源：Semantic Scholar
  • S2被引：0
  • 影响力被引：0
  • TLDR：This work designs a contrastive study that combines controlled quantitative experiments with paired trajectory analysis and consolidates observations into a taxonomy of three high-level categories and twelve skill-use modes, showing that skills work when noisy trajectories become procedural anchors that stabilize execution.
  • 副分类：evaluation
  • 待LLM分类：否
  • 标题中文：揭秘 Agent Skill：为何它们有效——直到失效
  • TLDR中文：本文设计了一项对比研究，结合受控定量实验与配对轨迹分析，并将观察结果归纳为一个包含三个高层类别与十二种 Skill 使用模式的分类法，表明当噪声轨迹转化为稳定执行的过程性锚点时，Skill 便会发挥作用。
  • 来源文件：
  • /inbox/tom/_candidates/2026-08-19-agent-rag-longcontext-candidates.json
  • [S2 enrich]