[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M758H2GER9T5QJ3ACMY7G5PB":3,"skill-same:01M758H2GER9T5QJ3ACMY7G5PB":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M758H2GER9T5QJ3ACMY7G5PB","omd-research-deep",null,"终极档深度调研:种子作者化多角度抓取 + council 分解 + 多轮缺口补挖,一条工具调用出整领域 grounded 报告。Trigger:\u002Fomd-research-deep、深度调研、deep research、整领域调研、调研到 grounded 底座、把这个领域研究透。Skip:轻量单点查证(omd MCP `omd_web`)\u002F 单问题综合(dag_research 默认档)\u002F 代码理解(dag-map)。","AbyssCN\u002Foh-my-dag","AbyssCN","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F43731064?v=4","01M758GF48B4J35SNWQZS4T8YM","01M758GF4YQNVGCZG82Y351MWV",44,"MIT","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Fclient-skills\u002Fomd-research-deep\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: omd-research-deep\ndescription: 终极档深度调研:种子作者化多角度抓取 + council 分解 + 多轮缺口补挖,一条工具调用出整领域 grounded 报告。Trigger:\u002Fomd-research-deep、深度调研、deep research、整领域调研、调研到 grounded 底座、把这个领域研究透。Skip:轻量单点查证(omd MCP `omd_web`)\u002F 单问题综合(dag_research 默认档)\u002F 代码理解(dag-map)。\n---\n\n# \u002Fomd-research-deep — 终极档深度调研\n\n调 omd MCP `dag_research`(可能带 `mcp__omd__` 前缀;未加载先 ToolSearch \"dag_research\"),\n开 **`super: true`** + **`rounds: 3`**。管线全在引擎内跑完才返回,**引擎计数,不问模型\"够了吗\"**:\n\n1. **种子作者化**(`super` 开的就是这个)— 模型把问题拆 3-4 个互补角度 query\n   (机制 \u002F 实践 \u002F 反面 \u002F 生态),各自独立检索并入语料。\n2. **council 分解** — conductor 按全部语料 author 领域专家 lens → L×V fanout →\n   per-lens 判优 → 综合 → judge panel → graft。(`council` 默认就是开的,不用传。)\n3. **多轮缺口补挖**(`rounds`,上限 4)— 轮间做 [模型缺口分析 + 确定性 probe:\n   引用集 − 已抓集 的缺料补抓],**无新增即提前停**;二轮起 challenger lens 只挖缺口不重答原题。\n\n## 用法\n\n```\ndag_research(question: \"\u003C研究问题>\", super: true, rounds: 3)\n```\n\n- **异步返回**(spawn detached 子进程):立即回 `runId`,用 `dag_status` 轮询到完成再取报告(真源 `src\u002Fmcp\u002Ftools\u002Fresearch.ts` 的 detached 分支)。\n- `summary` 进对话;**全文在 `reportPath`**(lens 冠军 + 逐轮缺口留痕 + 全部语料附录零丢失)。\n  关键决策**必须 `Read` 那个文件**,别只看 summary。\n- 转述纪律:结论 + 来源 URL + 哪轮缺口补出了什么;「语料未覆盖」的部分**如实说**。\n- 成本形状:种子×检索 + council + ≤N 轮,比默认档 `dag_research` 贵数倍\n  ——**真要挖透的领域才用**。\n\n## 前置与降级\n\n- **必须有 search provider**:`TAVILY_API_KEY` \u002F `ANYSEARCH_API_KEY` \u002F `SEARXNG_URL` 任一;`OMD_WEB_ORDER=searxng,anysearch,tavily` 覆盖默认 failover 序(默认 tavily→anysearch→searxng)。\n  没有 → 工具**响亮拒绝**(不会静默降级成\"看起来像调研的一段话\")。这是有意的:\n  没有 web 就没有调研。\n- **query 带「知乎\u002Fzhihu」→ 自动加一路 MediaCrawler 知乎站内搜**(需 `MEDIACRAWLER_URL` + 登录 cookie):分钟级,只用原 query 搜一次;挂了 warn 不断链。\n- **锚点文件 \u002F 显式种子 query 不在 MCP 面上**:`--anchor`(把已有设计笔记原样送进\n  groundTruth 之首)和 `--queries`(给死种子、不作者化)只有源码档的\n  `bun run scripts\u002Fdag-research.ts` 有。要这两个 → 直接跑脚本(需 omd 源码\u002F包目录),\n  或把锚点要点写进 `question` 正文。\n\n## 与既有能力的边界\n\n- `dag_research` **默认档**(不开 `super`)= 单问题综合,便宜得多 —— 大多数调研用它就够。\n- `omd_web` = 只抓不综合 · `omd_distill` = 吃已有料蒸洞察 · `dag-map` = codegraph 代码理解。\n- 轻量单点查证走客户端自带的 web 检索,别为一个事实起这条管线。\n- 领域法定源 RAG(会计\u002F法条类)不在此管线。\n","# \u002Fomd-research-deep — 终极档深度调研\n\n调 omd MCP `dag_research`(可能带 `mcp__omd__` 前缀;未加载先 ToolSearch \"dag_research\"),\n开 **`super: true`** + **`rounds: 3`**。管线全在引擎内跑完才返回,**引擎计数,不问模型\"够了吗\"**:\n\n1. **种子作者化**(`super` 开的就是这个)— 模型把问题拆 3-4 个互补角度 query\n   (机制 \u002F 实践 \u002F 反面 \u002F 生态),各自独立检索并入语料。\n2. **council 分解** — conductor 按全部语料 author 领域专家 lens → L×V fanout →\n   per-lens 判优 → 综合 → judge panel → graft。(`council` 默认就是开的,不用传。)\n3. **多轮缺口补挖**(`rounds`,上限 4)— 轮间做 [模型缺口分析 + 确定性 probe:\n   引用集 − 已抓集 的缺料补抓],**无新增即提前停**;二轮起 challenger lens 只挖缺口不重答原题。\n\n## 用法\n\n```\ndag_research(question: \"\u003C研究问题>\", super: true, rounds: 3)\n```\n\n- **异步返回**(spawn detached 子进程):立即回 `runId`,用 `dag_status` 轮询到完成再取报告(真源 `src\u002Fmcp\u002Ftools\u002Fresearch.ts` 的 detached 分支)。\n- `summary` 进对话;**全文在 `reportPath`**(lens 冠军 + 逐轮缺口留痕 + 全部语料附录零丢失)。\n  关键决策**必须 `Read` 那个文件**,别只看 summary。\n- 转述纪律:结论 + 来源 URL + 哪轮缺口补出了什么;「语料未覆盖」的部分**如实说**。\n- 成本形状:种子×检索 + council + ≤N 轮,比默认档 `dag_research` 贵数倍\n  ——**真要挖透的领域才用**。\n\n## 前置与降级\n\n- **必须有 search provider**:`TAVILY_API_KEY` \u002F `ANYSEARCH_API_KEY` \u002F `SEARXNG_URL` 任一;`OMD_WEB_ORDER=searxng,anysearch,tavily` 覆盖默认 failover 序(默认 tavily→anysearch→searxng)。\n  没有 → 工具**响亮拒绝**(不会静默降级成\"看起来像调研的一段话\")。这是有意的:\n  没有 web 就没有调研。\n- **query 带「知乎\u002Fzhihu」→ 自动加一路 MediaCrawler 知乎站内搜**(需 `MEDIACRAWLER_URL` + 登录 cookie):分钟级,只用原 query 搜一次;挂了 warn 不断链。\n- **锚点文件 \u002F 显式种子 query 不在 MCP 面上**:`--anchor`(把已有设计笔记原样送进\n  groundTruth 之首)和 `--queries`(给死种子、不作者化)只有源码档的\n  `bun run scripts\u002Fdag-research.ts` 有。要这两个 → 直接跑脚本(需 omd 源码\u002F包目录),\n  或把锚点要点写进 `question` 正文。\n\n## 与既有能力的边界\n\n- `dag_research` **默认档**(不开 `super`)= 单问题综合,便宜得多 —— 大多数调研用它就够。\n- `omd_web` = 只抓不综合 · `omd_distill` = 吃已有料蒸洞察 · `dag-map` = codegraph 代码理解。\n- 轻量单点查证走客户端自带的 web 检索,别为一个事实起这条管线。\n- 领域法定源 RAG(会计\u002F法条类)不在此管线。\n",{"name":5,"description":7},"main","624a2e346a5bcd50024a70081b3b5a7f8a80c72b","client-skills\u002Fomd-research-deep",2,"https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag","https:\u002F\u002Fgithub.com\u002FAbyssCN","2026-10-11 22:17:23",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add AbyssCN\u002Foh-my-dag --skill omd-research-deep",{"tool":36,"command":37},"bunx","bunx skills add AbyssCN\u002Foh-my-dag --skill omd-research-deep",{"tool":39,"command":40},"pnpm","pnpm dlx skills add AbyssCN\u002Foh-my-dag --skill omd-research-deep","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Fclient-skills\u002Fomd-research-deep\u002FSKILL.md",true,[],[45,51,57,63,70,77],{"skillNo":46,"name":47,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":48,"categories":49,"keywords":50,"publishedAt":18},"01M758HVAFZ52XFKX1VP0PD97F","skill-with-check","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fs1\u002Fskill-with-check\u002FSKILL.md",[],[],{"skillNo":52,"name":53,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":18},"01M758HRR8DJR96R16ZRTJZ5SC","skill-prose-bans","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fs1\u002Fskill-prose-bans\u002FSKILL.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":18},"01M758HPR7948QXXHKR2BT2PMN","skill-escalation-skill","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fs1\u002Fskill-escalation-skill\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":18},"01M758HM9488S9X0J803DV4XYD","bugfix-triage","修 bug 前的定位方法论 —— 最小复现、二分缩圈、区分\"测试错了\"与\"实现错了\"","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Ftemplates\u002Fpacks\u002Ftdd-bugfix\u002Fskills\u002Fbugfix-triage\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":18},"01M758HJ9V6ZK77GVRGZ0BHENS","ui-reviewer","Review rendered UI screenshots for hierarchy, spacing, readability, states, consistency, and visual slop.","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Fclient-skills\u002Fui-reviewer\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":18},"01M758HGQ636V2W83E7TTH3C80","sast","确定性 semgrep 静态扫描(SAST)+ 结构化报告。零 LLM 零成本,规则命中即事实。Trigger:\u002Fsast、静态扫描、semgrep、扫一下代码。","https:\u002F\u002Fgithub.com\u002FAbyssCN\u002Foh-my-dag\u002Fblob\u002Fmain\u002Fclient-skills\u002Fsast\u002FSKILL.md",[],[]]