{
  "schemaVersion": "bitshovel-editorial-review-v1",
  "id": "reading-quality-92-20260912",
  "signalId": "alphasignal-alphasignal-ai-news-what-8-100-trial-records-rev",
  "originalUrl": "https://arxiv.org/abs/2608.14036",
  "kind": "editorial-correction",
  "reviewedAt": "2026-09-12T13:30:20.540Z",
  "before": {
    "id": "alphasignal-alphasignal-ai-news-what-8-100-trial-records-rev",
    "state": "new",
    "lane": "workflow",
    "publishedAt": "2026-08-24T16:29:12.000Z",
    "observedAt": "2026-08-25T15:31:06.113Z",
    "originalUrl": "https://arxiv.org/abs/2608.14036",
    "relatedSources": [
      {
        "label": {
          "zh": "AlphaSignal 日报",
          "en": "AlphaSignal daily"
        },
        "url": "https://alphasignal.ai/news/what-8-100-trial-records-reveal-about-the-strengths-and-flaws-in-ai-agent-skills"
      }
    ],
    "source": {
      "zh": "arXiv 论文原文，经 AlphaSignal 日报进入视野",
      "en": "arXiv paper via the AlphaSignal daily"
    },
    "title": {
      "zh": "8,135 条试验记录拆解：Agent Skills 何时有效、为何失效",
      "en": "Demystifying agent skills: when they work, and until when"
    },
    "summary": {
      "zh": "论文用受控实验加轨迹配对分析回答三个问题：skills 什么时候帮得上忙、为什么起作用、在哪里失效。作者归一化 8,135 条试验记录、保留 238 个有效编码标签，归纳出三大类十二种 skill 使用模式：skills 的主要作用机制是把嘈杂的执行轨迹稳定成可复用的程序锚点（占 65.7% 的案例），匹配对比下比 Workflow Memory 高 6.06 分；跨框架复用与检索难度是主要失效来源。",
      "en": "The paper runs controlled experiments plus paired trajectory analysis over 8,135 normalized trial records (238 valid open-coded labels) to answer when skills help, why they work, and where they fail. Skills mostly act as procedural anchors that stabilize noisy execution (65.7% of cases), beating Workflow Memory by 6.06 points in matched comparisons; cross-framework reuse and retrieval difficulty are the main failure modes."
    },
    "why": {
      "zh": "Skills 已是 Claude Code 等工具的日常机制，但多数讨论停在'有没有用'。这篇给出可操作的失效边界（检索难度、跨框架复用），对决定哪些工作流值得沉淀成 skill 直接有用。",
      "en": "Skills are already everyday machinery in tools like Claude Code, but most discussion stops at 'do they help'. This maps the failure boundaries (retrieval difficulty, cross-framework reuse), which directly informs which workflows are worth distilling into skills."
    },
    "caveat": {
      "zh": "证据来自论文作者的受控基准实验，不是我们在生产环境的复测；基准上的程序锚点效应不保证迁移到任意私有工作流。",
      "en": "Evidence is the authors' controlled benchmark study — not proof from our production re-test; the procedural-anchoring effect on benchmarks may not transfer to arbitrary private workflows."
    },
    "timeline": [
      {
        "date": {
          "zh": "8 月 25 日",
          "en": "Aug 25"
        },
        "datetime": "2026-08-25T14:22:34.036Z",
        "summary": {
          "zh": "进入 BitShovel 雷达：AlphaSignal 每日 收录时可见 100 条。",
          "en": "Entered the BitShovel radar: 100 entries visible on AlphaSignal daily at collection."
        },
        "source": {
          "zh": "AlphaSignal 每日 快照",
          "en": "AlphaSignal daily snapshot"
        }
      }
    ],
    "followTarget": {
      "id": "page:arxiv-2608-14036",
      "label": {
        "zh": "Agent Skills 失效机理研究",
        "en": "Agent-skills failure-mode study"
      },
      "scope": {
        "zh": "论文修订、后续引用与复现结果",
        "en": "Revisions, citations, and replication results"
      },
      "url": "https://alphasignal.ai/news/what-8-100-trial-records-reveal-about-the-strengths-and-flaws-in-ai-agent-skills",
      "updateMode": "verified-change"
    },
    "ingestion": {
      "mode": "editorial",
      "adapter": "syndication-feed",
      "evidenceClass": "institutional-thesis",
      "observationRef": "source-observations/history/syndication-feeds.2026-08-25T15-31-06Z.snapshot.json#feed:alphasignal-daily",
      "changeRef": "source-observations/history/syndication-feed-changes.2026-08-25T15-31-06Z.snapshot.json#feed:alphasignal-daily"
    },
    "verifiedAt": "2026-09-01T22:00:14.205Z",
    "retireReceipt": {
      "decision": "retire-candidate",
      "basis": "silent",
      "lateral7d": 0,
      "vertical7d": 0,
      "decidedAt": "2026-09-01T22:01:26.477Z"
    },
    "__retiredAt": "2026-09-01T22:01:26.477Z"
  },
  "after": {
    "title": {
      "zh": "Agent Skills 为什么有用：把经验变成步骤，还要适配当前任务",
      "en": "Why Agent Skills help: reusable steps still need adaptation"
    },
    "summary": {
      "zh": "Skills 把以往执行过程压缩成可复用步骤，帮助 Agent 少走重复弯路。这篇研究同时说明：有合适的 Skill、找得到它和最终完成任务，是不同的问题。",
      "en": "Skills turn prior execution experience into reusable procedures. The study separates having useful guidance, finding it and actually completing the task."
    },
    "why": {
      "zh": "对照同一批经验的不同组织方式，有助于理解步骤、检索和环境适配各自的作用。",
      "en": "Comparing representations of the same experience helps distinguish procedural guidance, retrieval and adaptation."
    },
    "caveat": {
      "zh": "这是有限模型与终端任务的研究，不是所有 Agent 的通用结论；机制分类来自抽样和配对分析，使用精确率不等于任务成功率。",
      "en": "This is a study of limited models and terminal tasks, not proof of general gains for all agents. Mechanism labels use sampling and paired analysis; usage precision is not task success."
    }
  },
  "note": {
    "zh": "说明轨迹分类的抽样范围与配对分析，区分 Skill 使用精确率和任务成功率。 保留事件、发现和收录日期。",
    "en": "Clarified sampled taxonomy and paired analysis, and separated skill-use precision from task success. Event, discovery and collection dates remain unchanged."
  },
  "withdrawnSources": [],
  "withdrawnTimeline": [],
  "evidence": [
    {
      "type": "retained-official-page",
      "label": {
        "zh": "arXiv 论文原文，经 AlphaSignal 日报进入视野",
        "en": "arXiv paper via the AlphaSignal daily"
      },
      "url": "https://arxiv.org/abs/2608.14036",
      "fetchedAt": "2026-09-12T13:16:16.669Z",
      "path": "app/_content/editorial-reviews/evidence/reading-quality-91-95-bd50251d560a-20260912.json",
      "bytes": 508,
      "sha256": "ab17f8ea5162361bc062b99dbb57da87385ecf18f44410f617a47d45802a0379"
    },
    {
      "type": "retained-official-page",
      "label": {
        "zh": "Skills 研究论文全文",
        "en": "Full Skills study"
      },
      "url": "https://arxiv.org/html/2608.14036v1",
      "fetchedAt": "2026-09-12T13:24:23.217Z",
      "path": "app/_content/editorial-reviews/evidence/reading-quality-91-95-871d7d421629-20260912.json",
      "bytes": 515,
      "sha256": "518e7dba111c2619aebcf2efc127bef2828a7c29f09468f8878c1d543befa432"
    }
  ],
  "integrity": "6c111eaf923addeb3b7f29a12c471ced8fac37fa48a4516369a2771fa771f705"
}