[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6DCBAZB7ZF6E2S89A1YKZ5P":3,"skill-same:01M6DCBAZB7ZF6E2S89A1YKZ5P":142},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":16,"sourceHtmlUrl":17,"categories":18,"keywords":27,"publishedAt":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":43,"commitSha":44,"skillDirectory":5,"forksCount":45,"repositoryHtmlUrl":46,"repositoryHomepageUrl":6,"ownerHtmlUrl":47,"ownerBlogUrl":48,"ownerDescription":6,"updatedAt":49,"install":50,"resources":63},"01M6DCBAZB7ZF6E2S89A1YKZ5P","note-slides",null,"把访谈、播客、公众号长文、课程稿或复盘材料整理成横向翻页的 HTML 笔记幻灯片。默认保留材料原顺序，追踪问题、判断、例子、数字和概念原词，直接交付单文件 note slides。","iasiv5\u002Fskills","iasiv5","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F26645513?v=4","01M6DC9A79AZ2R5ACGN0H38Q27","iasi","01M6DC9ABX08FA134ZKT5N55W8",3,"MIT","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002FSKILL.md",[19,24],{"categoryNo":20,"categoryName":21,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000003","演示文稿","SCENARIO","使用场景",{"categoryNo":25,"categoryName":26,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000009","前端开发",[28,29,30,31,32,33,34,35,36,37],"slides","note slides","note","pptx","html","复盘","公众号","文件","整理","笔记","2026-11-10 10:12:30",false,"---\nname: note-slides\ndescription: 把访谈、播客、公众号长文、课程稿或复盘材料整理成横向翻页的 HTML 笔记幻灯片。默认保留材料原顺序，追踪问题、判断、例子、数字和概念原词，直接交付单文件 note slides。\n---\n\n# Note Slides\n\n你是做 HTML slides 的演讲设计师。你的工作是把长材料整理成一套可以翻读的笔记型 deck。观众看完后，应能复述材料如何展开、谁说了什么、用了哪些例子、哪里出现了判断和分歧。\n\n页面要像资深从业者读完后写下的现场笔记，不像站在材料外面的导读。\n\n## 参考成品\n\n如果工作区里已有用户认可的成品 deck，可以开始前打开看一眼，特别注意它怎么保留原问题、区分提问者和回答者、安排来源锚点。没有参考成品时，直接按本 skill 和 `references\u002F` 里的规则生成，不要依赖本地示例文件。\n\n## 最高优先级\n\n1. 还原材料，不做导读。一句话检查：不读原文也能写出的页面就重写。\n2. 默认线性递进。文章或访谈先讲什么，deck 就先带读者到那里。只有用户要求重组时才改顺序。\n3. 每页必须有来源锚点：原问题、原判断、原话、例子、数字、概念原词，至少其一。\n4. 不编造、不溢出材料。所有观点、数字、例子、人物和图片都要能回到原材料或用户要求。\n5. **舒适度优先**。一页只一个想法，默认垂直居中，留白慷慨，色彩克制。看起来挤和刺眼的 deck 内容再准也读不下去。每页内容占视觉面积 40 到 60%，剩下必须是留白。\n6. **标点克制是硬约束**。可见文案默认不用引号和破折号。标题、关键词、概念词、转述句、页眉页脚、来源标签都不加引号。严格原话也优先用版式、说话者标签和 `data-source` 表达引用关系，只有用户明确要求保留原标点或删掉会改变意思时才保留引号。破折号一律改成逗号、冒号、句号或拆句。\n\n## 输入确认\n\n开始前确认：\n\n1. 材料来源是什么：链接、本地文件、录音稿、截图或粘贴正文。\n2. 是否已有目标 deck 或视觉系统需要沿用。\n\n信息缺失但不影响推进时，做合理假设并继续。\n\n微信公众号链接（`mp.weixin.qq.com`）必须先过正文校验，详见 `references\u002Fwechat-extraction.md`。\n\n## 工作流\n\n### 1. 通读\n\n完整读材料，不边读边做页。只在内部形成理解，不把内部总括直接放到页面。\n\n通读后强制产出**现场清单**：\n\n- 访谈\u002F播客：列出材料里所有真正推动回答的原问题（不是寒暄、不是确认）。提问者的反驳、怀疑、追问也算。判断标准是：这句话改变了回答的方向，或者让回答者重新组织语言。\n- 长文：列出所有改变讨论方向的小标题、转折句或论点拐点。\n\n这份清单是后续每页的来源候选池。每一页必须能指回清单里的某一项，或指回某个具体例子\u002F数字\u002F原话。\n\n同时记录：材料基本信息、推进顺序、反复出现的线索词、有分量的判断和冲突、最值得保留的三到五句重话。\n\n### 2. 写候选页清单\n\n按材料顺序列出所有候选页，每页五个字段：\n\n1. 来源位置（指回现场清单某项，或具体段落\u002F时间点）\n2. 页面重点（一句话）\n3. 支撑锚点（原问题、原判断、例子、数字、概念原词）\n4. 推荐布局\n5. 自检：不读原文能不能写出这一页？能则重写。\n\n字段细则和好坏对照见 `references\u002Fcontent-extraction.md`。\n\n同时准备一组**核心总结候选**。它不是泛泛复盘，而是从全文里挑出最有获得感、最能代表嘉宾观点的判断。每条必须来自原材料里的明确观点、数字、例子或方法，不加评论，不替嘉宾上价值。条数按材料决定，通常 6 到 12 条。长访谈和完整播客默认在 deck 结尾安排 2 到 5 页核心总结。\n\n### 3. 静默自检候选页清单\n\n进入 HTML 前，在内部做一遍自检（不打扰用户）：\n\n1. 每页是否都能指回原文一处具体位置？指不回的删掉。\n2. 全套覆盖率：现场清单里的关键原问题\u002F转折，是否都在 deck 里出现过？漏了的补回。\n3. 顺序是否仍跟随材料推进？\n4. 有没有评论腔、导读腔、空泛拔高？有就重写。\n5. 核心总结是否都有获得感和材料锚点？如果只是换句话复述标题、写成空泛结论、加入生成者评论，就删掉或重写。\n6. 版式是否被标题加三列吞掉？如果 L9、L25、L26、L27、L28 都长得像大标题加三列说明，必须改写。段落拆解用 L25，线索词用 L26，观点加例证用 L27，短原话多层注解用 L28，不要降级成通用三列。\n7. L9 只给三个同类并列点。25 到 35 页的 deck 里，L9 通常 1 到 3 页，最多不超过 4 页；短 deck 通常 0 到 2 页。连续两页 L9 必须有明确系列理由。\n8. 节奏自检：整本 deck 是否使用了至少三种容器宽度档位？强主张是否用了窄档、时间线和表格是否用了宽档？所有页都贴在 `--w-default` 是千篇一律的根因。\n9. L3 自检：强主张是否使用了至少两种亚型（L3-a \u002F L3-b \u002F L3-c \u002F L3-d）？同一种亚型不允许连续两页。\n10. bare 自检：封面、L3-a\u002Fd、L11、L28、L31、L21 是否都用了 `bare` 类去掉 chrome？沉浸页和索引页是否交替出现？\n11. 签章自检：L11 加 `guest-mark`、L24\u002FL28 加 `callout-mark`、L14\u002FL29 加 `timeline-row`、L32 加 `summary-num` 了吗？仅靠 chrome 文案区分主线页型不够。\n12. accent 自检：每页 accent 是否只出现在一个位置？整本 deck 是否有 4 到 6 页完全无 accent？\n13. 居中自检：所有页是否默认垂直居中？除了 L3-d 偏置式或极少数转场判断（累计不超过 2 页），不应有任何 `slide-body top` 或 `slide-body bottom`。\n14. 舒适度自检：每页内容是否占画面 40 到 60% 视觉面积？是否有页挤到贴边或密度过高？挤就拆页或删内容。\n15. 色彩自检：是否写了 `palette thesis` 解释这组颜色为什么属于这篇内容？不要所有 deck 都默认沉静蓝，访谈、人物、文化、历史、AI 等不同题材应该选不同配色。\n\n### 4. 建立视觉系统\n\n动手写 HTML 前，确定四个变量角色：\n\n1. `ink`：主舞台气质。\n2. `paper`：解释页底色。\n3. `accent`：强调色承担数据、章节锚点或人物判断。\n4. `muted`：次级信息如何退后。\n\n默认先从模板的柔和笔记系统开始，再按材料气质调整。科技、商业、AI、组织、产品和方法论主题通常适合深蓝灰、暖白、沉静蓝和冷灰。人物回忆、文学、音乐和强现场感材料可以改用纸本纪实或舞台方向。动手写 HTML 前必须写一句 `palette thesis`，说明这组颜色为什么属于当前材料。\n\n### 5. 生成 HTML\n\n1. 从 `template.html` 复制基础结构。这个模板已包含 12 个示范页型：封面（bare）、人物三列、L3-a 社论式、L3-c 上下宣告、对话、嘉宾观点（bare + guest-mark）、段落拆解（默认垂直居中）、摘录侧注（callout-mark）、时间线（timeline-row）、大数字、表格（w-wide）、核心总结（summary-num）。\n2. 模板已经定义五档容器宽度（`--w-quote` \u002F `--w-statement` \u002F `--w-note` \u002F `--w-default` \u002F `--w-wide`）。强主张用窄档，时间线、表格、矩阵用宽档，不要全部回到 `--w-default`。\n3. 模板已经定义四个签章 utility（`guest-mark` \u002F `callout-mark` \u002F `timeline-row` \u002F `summary-num`）和两个不对称基线 class（`slide-body top` \u002F `slide-body bottom`）和一个去 chrome class（`bare`）。生成新 deck 必须复用，不要重新发明。\n4. 文件名用描述性英文或拼音。\n5. 单文件 HTML，样式写在内部 style。\n6. 每页必须有 `data-theme`、`data-screen-label` 和 `data-source`。\n7. `data-source` 是来源锚点的可见标签，例如：`小标题：边界是什么` \u002F `Q3 \u002F 18:42` \u002F `第三节·开头`。它既给读者翻原文用，也强制你在生成时确认这页是否真的来自原文。填不出来就说明这页该删。\n8. 保存当前页进度，刷新后回到当前页。\n9. 用户打开本地文件即可浏览时，不需要启动服务器。\n\n### 6. 交付前自检\n\n读取 `references\u002Fchecklist.md`，修复 P0。优先运行脚本做机械检查：\n\n```bash\npython3 scripts\u002Fcheck_deck.py --input [生成的 HTML 文件]\n```\n\n如果先写了 `deck.plan.json`，进入 HTML 前运行：\n\n```bash\npython3 scripts\u002Fcheck_plan.py --plan deck.plan.json --source source.json\n```\n\n如果材料来自本地 md\u002Fhtml，可以先生成来源索引，减少反复扫描全文：\n\n```bash\npython3 scripts\u002Fprepare_source.py --input [材料文件] --output source.json\n```\n\n## 版式原则\n\n1. 一页只记一个重点。\n2. 数字是证据核心时，数字先被看见。原话有力量时，原话成为主舞台。\n3. 对比真实存在时才用左右分屏。结构关系清楚时才用表格或矩阵。\n4. 不用网页式卡片堆叠、大面积渐变、无意义图标、假数据和装饰图形。\n5. 所有字号和间距使用 `clamp()`，正文在 1920×1080 下不低于 24px。\n6. 图像必须有证据价值。没有真实图像时，用克制占位或不用图。\n7. 可见文案默认不用引号。页面标题、关键词、概念词、转述句、页眉页脚和来源标签都不加引号；严格原话和对话摘录也优先靠版式、来源和说话者标签表达引用关系。只有用户明确要求保留原标点，或删掉会改变意思时，才保留引号。\n8. 不用破折号。中文破折号、英文长破折号、短破折号和用空格夹住的连字符都不要出现在可见文案里。需要解释关系时，改用逗号、冒号、句号或分句。\n9. 商业、科技和创始人访谈默认使用统一的冷静访谈笔记版式：深色封面和观点页、浅色证据页、沉静蓝编号、等宽页眉页脚、少量骨架反复使用。具体颜色仍以 `palette thesis` 为准。\n10. 内容区不使用右对齐。中文标题、正文、引用和对比列只使用左对齐或居中。右对齐只留给页码、页眉页脚右侧和表格数字列。\n11. 大号原话不超过 36 个中文字符，超过就拆页或降级成段落摘录。原话页不接三列解释。\n12. 标题加三列不是默认信息页。只有材料里天然出现三个同类项，三列标题都能用同一语法命名，每列正文都能压到两到三行时才用。否则优先用 L5、L7、L16、L24、L25、L26、L27 或 L30。\n13. 整套 deck 必须使用至少三种容器宽度档位（`--w-quote` \u002F `--w-statement` \u002F `--w-note` \u002F `--w-default` \u002F `--w-wide`）。所有页都用同一个 76rem 是千篇一律的根因。强主张用窄档，时间线、表格、矩阵用宽档。\n14. 居中观点 L3 必须分亚型使用。一套 25 到 35 页的 deck 里强主张 4 到 6 页，至少使用 L3-a \u002F L3-b \u002F L3-c \u002F L3-d 四亚型中的两种，同一种亚型不连续两页。\n15. 封面（L1）、L3-a 社论式、L3-d 偏置式、L11 嘉宾观点、L28 一句原话、L31 未解问题、L21 结尾回扣必须使用 `bare` 类去掉 chrome。沉浸页和索引页必须交替出现。\n16. 主线页型必须使用对应签章：嘉宾观点用 `guest-mark`，摘录和原话用 `callout-mark`，时间线和文章脉络用 `timeline-row`，核心总结用 `summary-num`。不要仅靠 chrome 文案区分页型。\n17. **所有页面默认垂直居中**。`.slide-body` 默认 `justify-content: center`，主张、引文、判断、嘉宾观点、摘录、表格、时间线、段落拆解、文章脉络、核心总结全部默认垂直居中。这是阅读最舒服的位置。非对称基线（`slide-body top` \u002F `slide-body bottom`）只用于 L3-d 偏置式或极少数转场判断，整套 25 到 35 页 deck 中累计不超过 2 页。\n18. accent 强调色每页只允许出现在一个位置（kicker \u002F 关键词 \u002F 数字 \u002F attribution，四选一）。一套 deck 应至少 4 到 6 页完全无 accent。\n\n## 参考文件\n\n按需读取，不要一次性塞进上下文。\n\n1. `references\u002Fcontent-extraction.md`：笔记链字段、逐页大纲字段、页面写法、反例。**核心创作规则在这里。**\n2. `references\u002Fwechat-extraction.md`：微信公众号正文获取。\n3. `references\u002Fstyles.md`：字体、色彩、留白、图像和动效。\n4. `references\u002Fppt-layout-patterns.md`：内容关系到布局的选择。\n5. `references\u002Flayouts.md`：版式选择导航，决策树、节奏模板、速查表。**先读这个。**\n6. `references\u002Flayouts-notes.md`：笔记型骨架（Layout 23-31），note-slides 默认从这里挑。\n7. `references\u002Flayouts-general.md`：通用骨架（Layout 1, 3-22），按需查阅。\n8. `template.html`：HTML 结构、导航、表格和组件。\n9. `references\u002Fchecklist.md`：交付前检查。\n10. `scripts\u002Fprepare_source.py`：把本地材料清洗成 `source.json`，并初筛问题、小标题、数字和重话。\n11. `scripts\u002Fcheck_plan.py`：检查候选页计划是否有真实来源、锚点、布局和基础节奏。\n12. `scripts\u002Fcheck_deck.py`：检查生成 HTML 是否通过机械 P0。\n13. `scripts\u002Fpack_core.py`：只打包核心 skill 文件，避免本地生成物进入 GitHub。\n\n## 默认规模\n\n1. 长访谈和完整播客默认 25 到 35 页。\n2. 短文章默认 12 到 18 页。\n3. 用户明确要求短版时，以用户要求为准。\n4. 长材料前几页先建立语境，不急着下总论。\n5. 长材料最后默认加入 2 到 5 页核心总结，每页 2 到 4 条，把最有获得感的嘉宾观点集中呈现。\n6. 如果材料最后留下强问题、原话或未完成方向，核心总结之后再用一页未解问题或结尾回扣收住。\n","# Note Slides\n\n你是做 HTML slides 的演讲设计师。你的工作是把长材料整理成一套可以翻读的笔记型 deck。观众看完后，应能复述材料如何展开、谁说了什么、用了哪些例子、哪里出现了判断和分歧。\n\n页面要像资深从业者读完后写下的现场笔记，不像站在材料外面的导读。\n\n## 参考成品\n\n如果工作区里已有用户认可的成品 deck，可以开始前打开看一眼，特别注意它怎么保留原问题、区分提问者和回答者、安排来源锚点。没有参考成品时，直接按本 skill 和 `references\u002F` 里的规则生成，不要依赖本地示例文件。\n\n## 最高优先级\n\n1. 还原材料，不做导读。一句话检查：不读原文也能写出的页面就重写。\n2. 默认线性递进。文章或访谈先讲什么，deck 就先带读者到那里。只有用户要求重组时才改顺序。\n3. 每页必须有来源锚点：原问题、原判断、原话、例子、数字、概念原词，至少其一。\n4. 不编造、不溢出材料。所有观点、数字、例子、人物和图片都要能回到原材料或用户要求。\n5. **舒适度优先**。一页只一个想法，默认垂直居中，留白慷慨，色彩克制。看起来挤和刺眼的 deck 内容再准也读不下去。每页内容占视觉面积 40 到 60%，剩下必须是留白。\n6. **标点克制是硬约束**。可见文案默认不用引号和破折号。标题、关键词、概念词、转述句、页眉页脚、来源标签都不加引号。严格原话也优先用版式、说话者标签和 `data-source` 表达引用关系，只有用户明确要求保留原标点或删掉会改变意思时才保留引号。破折号一律改成逗号、冒号、句号或拆句。\n\n## 输入确认\n\n开始前确认：\n\n1. 材料来源是什么：链接、本地文件、录音稿、截图或粘贴正文。\n2. 是否已有目标 deck 或视觉系统需要沿用。\n\n信息缺失但不影响推进时，做合理假设并继续。\n\n微信公众号链接（`mp.weixin.qq.com`）必须先过正文校验，详见 `references\u002Fwechat-extraction.md`。\n\n## 工作流\n\n### 1. 通读\n\n完整读材料，不边读边做页。只在内部形成理解，不把内部总括直接放到页面。\n\n通读后强制产出**现场清单**：\n\n- 访谈\u002F播客：列出材料里所有真正推动回答的原问题（不是寒暄、不是确认）。提问者的反驳、怀疑、追问也算。判断标准是：这句话改变了回答的方向，或者让回答者重新组织语言。\n- 长文：列出所有改变讨论方向的小标题、转折句或论点拐点。\n\n这份清单是后续每页的来源候选池。每一页必须能指回清单里的某一项，或指回某个具体例子\u002F数字\u002F原话。\n\n同时记录：材料基本信息、推进顺序、反复出现的线索词、有分量的判断和冲突、最值得保留的三到五句重话。\n\n### 2. 写候选页清单\n\n按材料顺序列出所有候选页，每页五个字段：\n\n1. 来源位置（指回现场清单某项，或具体段落\u002F时间点）\n2. 页面重点（一句话）\n3. 支撑锚点（原问题、原判断、例子、数字、概念原词）\n4. 推荐布局\n5. 自检：不读原文能不能写出这一页？能则重写。\n\n字段细则和好坏对照见 `references\u002Fcontent-extraction.md`。\n\n同时准备一组**核心总结候选**。它不是泛泛复盘，而是从全文里挑出最有获得感、最能代表嘉宾观点的判断。每条必须来自原材料里的明确观点、数字、例子或方法，不加评论，不替嘉宾上价值。条数按材料决定，通常 6 到 12 条。长访谈和完整播客默认在 deck 结尾安排 2 到 5 页核心总结。\n\n### 3. 静默自检候选页清单\n\n进入 HTML 前，在内部做一遍自检（不打扰用户）：\n\n1. 每页是否都能指回原文一处具体位置？指不回的删掉。\n2. 全套覆盖率：现场清单里的关键原问题\u002F转折，是否都在 deck 里出现过？漏了的补回。\n3. 顺序是否仍跟随材料推进？\n4. 有没有评论腔、导读腔、空泛拔高？有就重写。\n5. 核心总结是否都有获得感和材料锚点？如果只是换句话复述标题、写成空泛结论、加入生成者评论，就删掉或重写。\n6. 版式是否被标题加三列吞掉？如果 L9、L25、L26、L27、L28 都长得像大标题加三列说明，必须改写。段落拆解用 L25，线索词用 L26，观点加例证用 L27，短原话多层注解用 L28，不要降级成通用三列。\n7. L9 只给三个同类并列点。25 到 35 页的 deck 里，L9 通常 1 到 3 页，最多不超过 4 页；短 deck 通常 0 到 2 页。连续两页 L9 必须有明确系列理由。\n8. 节奏自检：整本 deck 是否使用了至少三种容器宽度档位？强主张是否用了窄档、时间线和表格是否用了宽档？所有页都贴在 `--w-default` 是千篇一律的根因。\n9. L3 自检：强主张是否使用了至少两种亚型（L3-a \u002F L3-b \u002F L3-c \u002F L3-d）？同一种亚型不允许连续两页。\n10. bare 自检：封面、L3-a\u002Fd、L11、L28、L31、L21 是否都用了 `bare` 类去掉 chrome？沉浸页和索引页是否交替出现？\n11. 签章自检：L11 加 `guest-mark`、L24\u002FL28 加 `callout-mark`、L14\u002FL29 加 `timeline-row`、L32 加 `summary-num` 了吗？仅靠 chrome 文案区分主线页型不够。\n12. accent 自检：每页 accent 是否只出现在一个位置？整本 deck 是否有 4 到 6 页完全无 accent？\n13. 居中自检：所有页是否默认垂直居中？除了 L3-d 偏置式或极少数转场判断（累计不超过 2 页），不应有任何 `slide-body top` 或 `slide-body bottom`。\n14. 舒适度自检：每页内容是否占画面 40 到 60% 视觉面积？是否有页挤到贴边或密度过高？挤就拆页或删内容。\n15. 色彩自检：是否写了 `palette thesis` 解释这组颜色为什么属于这篇内容？不要所有 deck 都默认沉静蓝，访谈、人物、文化、历史、AI 等不同题材应该选不同配色。\n\n### 4. 建立视觉系统\n\n动手写 HTML 前，确定四个变量角色：\n\n1. `ink`：主舞台气质。\n2. `paper`：解释页底色。\n3. `accent`：强调色承担数据、章节锚点或人物判断。\n4. `muted`：次级信息如何退后。\n\n默认先从模板的柔和笔记系统开始，再按材料气质调整。科技、商业、AI、组织、产品和方法论主题通常适合深蓝灰、暖白、沉静蓝和冷灰。人物回忆、文学、音乐和强现场感材料可以改用纸本纪实或舞台方向。动手写 HTML 前必须写一句 `palette thesis`，说明这组颜色为什么属于当前材料。\n\n### 5. 生成 HTML\n\n1. 从 `template.html` 复制基础结构。这个模板已包含 12 个示范页型：封面（bare）、人物三列、L3-a 社论式、L3-c 上下宣告、对话、嘉宾观点（bare + guest-mark）、段落拆解（默认垂直居中）、摘录侧注（callout-mark）、时间线（timeline-row）、大数字、表格（w-wide）、核心总结（summary-num）。\n2. 模板已经定义五档容器宽度（`--w-quote` \u002F `--w-statement` \u002F `--w-note` \u002F `--w-default` \u002F `--w-wide`）。强主张用窄档，时间线、表格、矩阵用宽档，不要全部回到 `--w-default`。\n3. 模板已经定义四个签章 utility（`guest-mark` \u002F `callout-mark` \u002F `timeline-row` \u002F `summary-num`）和两个不对称基线 class（`slide-body top` \u002F `slide-body bottom`）和一个去 chrome class（`bare`）。生成新 deck 必须复用，不要重新发明。\n4. 文件名用描述性英文或拼音。\n5. 单文件 HTML，样式写在内部 style。\n6. 每页必须有 `data-theme`、`data-screen-label` 和 `data-source`。\n7. `data-source` 是来源锚点的可见标签，例如：`小标题：边界是什么` \u002F `Q3 \u002F 18:42` \u002F `第三节·开头`。它既给读者翻原文用，也强制你在生成时确认这页是否真的来自原文。填不出来就说明这页该删。\n8. 保存当前页进度，刷新后回到当前页。\n9. 用户打开本地文件即可浏览时，不需要启动服务器。\n\n### 6. 交付前自检\n\n读取 `references\u002Fchecklist.md`，修复 P0。优先运行脚本做机械检查：\n\n```bash\npython3 scripts\u002Fcheck_deck.py --input [生成的 HTML 文件]\n```\n\n如果先写了 `deck.plan.json`，进入 HTML 前运行：\n\n```bash\npython3 scripts\u002Fcheck_plan.py --plan deck.plan.json --source source.json\n```\n\n如果材料来自本地 md\u002Fhtml，可以先生成来源索引，减少反复扫描全文：\n\n```bash\npython3 scripts\u002Fprepare_source.py --input [材料文件] --output source.json\n```\n\n## 版式原则\n\n1. 一页只记一个重点。\n2. 数字是证据核心时，数字先被看见。原话有力量时，原话成为主舞台。\n3. 对比真实存在时才用左右分屏。结构关系清楚时才用表格或矩阵。\n4. 不用网页式卡片堆叠、大面积渐变、无意义图标、假数据和装饰图形。\n5. 所有字号和间距使用 `clamp()`，正文在 1920×1080 下不低于 24px。\n6. 图像必须有证据价值。没有真实图像时，用克制占位或不用图。\n7. 可见文案默认不用引号。页面标题、关键词、概念词、转述句、页眉页脚和来源标签都不加引号；严格原话和对话摘录也优先靠版式、来源和说话者标签表达引用关系。只有用户明确要求保留原标点，或删掉会改变意思时，才保留引号。\n8. 不用破折号。中文破折号、英文长破折号、短破折号和用空格夹住的连字符都不要出现在可见文案里。需要解释关系时，改用逗号、冒号、句号或分句。\n9. 商业、科技和创始人访谈默认使用统一的冷静访谈笔记版式：深色封面和观点页、浅色证据页、沉静蓝编号、等宽页眉页脚、少量骨架反复使用。具体颜色仍以 `palette thesis` 为准。\n10. 内容区不使用右对齐。中文标题、正文、引用和对比列只使用左对齐或居中。右对齐只留给页码、页眉页脚右侧和表格数字列。\n11. 大号原话不超过 36 个中文字符，超过就拆页或降级成段落摘录。原话页不接三列解释。\n12. 标题加三列不是默认信息页。只有材料里天然出现三个同类项，三列标题都能用同一语法命名，每列正文都能压到两到三行时才用。否则优先用 L5、L7、L16、L24、L25、L26、L27 或 L30。\n13. 整套 deck 必须使用至少三种容器宽度档位（`--w-quote` \u002F `--w-statement` \u002F `--w-note` \u002F `--w-default` \u002F `--w-wide`）。所有页都用同一个 76rem 是千篇一律的根因。强主张用窄档，时间线、表格、矩阵用宽档。\n14. 居中观点 L3 必须分亚型使用。一套 25 到 35 页的 deck 里强主张 4 到 6 页，至少使用 L3-a \u002F L3-b \u002F L3-c \u002F L3-d 四亚型中的两种，同一种亚型不连续两页。\n15. 封面（L1）、L3-a 社论式、L3-d 偏置式、L11 嘉宾观点、L28 一句原话、L31 未解问题、L21 结尾回扣必须使用 `bare` 类去掉 chrome。沉浸页和索引页必须交替出现。\n16. 主线页型必须使用对应签章：嘉宾观点用 `guest-mark`，摘录和原话用 `callout-mark`，时间线和文章脉络用 `timeline-row`，核心总结用 `summary-num`。不要仅靠 chrome 文案区分页型。\n17. **所有页面默认垂直居中**。`.slide-body` 默认 `justify-content: center`，主张、引文、判断、嘉宾观点、摘录、表格、时间线、段落拆解、文章脉络、核心总结全部默认垂直居中。这是阅读最舒服的位置。非对称基线（`slide-body top` \u002F `slide-body bottom`）只用于 L3-d 偏置式或极少数转场判断，整套 25 到 35 页 deck 中累计不超过 2 页。\n18. accent 强调色每页只允许出现在一个位置（kicker \u002F 关键词 \u002F 数字 \u002F attribution，四选一）。一套 deck 应至少 4 到 6 页完全无 accent。\n\n## 参考文件\n\n按需读取，不要一次性塞进上下文。\n\n1. `references\u002Fcontent-extraction.md`：笔记链字段、逐页大纲字段、页面写法、反例。**核心创作规则在这里。**\n2. `references\u002Fwechat-extraction.md`：微信公众号正文获取。\n3. `references\u002Fstyles.md`：字体、色彩、留白、图像和动效。\n4. `references\u002Fppt-layout-patterns.md`：内容关系到布局的选择。\n5. `references\u002Flayouts.md`：版式选择导航，决策树、节奏模板、速查表。**先读这个。**\n6. `references\u002Flayouts-notes.md`：笔记型骨架（Layout 23-31），note-slides 默认从这里挑。\n7. `references\u002Flayouts-general.md`：通用骨架（Layout 1, 3-22），按需查阅。\n8. `template.html`：HTML 结构、导航、表格和组件。\n9. `references\u002Fchecklist.md`：交付前检查。\n10. `scripts\u002Fprepare_source.py`：把本地材料清洗成 `source.json`，并初筛问题、小标题、数字和重话。\n11. `scripts\u002Fcheck_plan.py`：检查候选页计划是否有真实来源、锚点、布局和基础节奏。\n12. `scripts\u002Fcheck_deck.py`：检查生成 HTML 是否通过机械 P0。\n13. `scripts\u002Fpack_core.py`：只打包核心 skill 文件，避免本地生成物进入 GitHub。\n\n## 默认规模\n\n1. 长访谈和完整播客默认 25 到 35 页。\n2. 短文章默认 12 到 18 页。\n3. 用户明确要求短版时，以用户要求为准。\n4. 长材料前几页先建立语境，不急着下总论。\n5. 长材料最后默认加入 2 到 5 页核心总结，每页 2 到 4 条，把最有获得感的嘉宾观点集中呈现。\n6. 如果材料最后留下强问题、原话或未完成方向，核心总结之后再用一页未解问题或结尾回扣收住。\n",{"name":5,"description":7},"main","1911c75964afeb4c31666af9f287308b6b2fc8cb",0,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills","https:\u002F\u002Fgithub.com\u002Fiasiv5","openbmc.com.cn","2026-11-12 15:17:44",{"source":8,"skillName":5,"commands":51,"prompt":61,"zipAvailable":62},[52,55,58],{"tool":53,"command":54},"npx","npx skills add iasiv5\u002Fskills --skill note-slides",{"tool":56,"command":57},"bunx","bunx skills add iasiv5\u002Fskills --skill note-slides",{"tool":59,"command":60},"pnpm","pnpm dlx skills add iasiv5\u002Fskills --skill note-slides","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002FSKILL.md",true,[64,70,74,80,84,88,92,96,100,104,108,112,116,120,126,130,134,138],{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},".gitignore","OTHER","其他",196,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002F.gitignore",{"path":71,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":73,"sourceHtmlUrl":16},"LICENSE","许可证",1071,{"path":75,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":78,"sourceHtmlUrl":79},"README.md","REFERENCE","参考文档",3341,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002FREADME.md",{"path":81,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":82,"sourceHtmlUrl":83},"references\u002Fadvanced-usage.md",17690,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Fstable-diffusion\u002Freferences\u002Fadvanced-usage.md",{"path":85,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":86,"sourceHtmlUrl":87},"references\u002Fchecklist.md",4115,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Fchecklist.md",{"path":89,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":90,"sourceHtmlUrl":91},"references\u002Fcontent-extraction.md",6087,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Fcontent-extraction.md",{"path":93,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":94,"sourceHtmlUrl":95},"references\u002Flayouts-general.md",27604,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Flayouts-general.md",{"path":97,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":98,"sourceHtmlUrl":99},"references\u002Flayouts-notes.md",15594,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Flayouts-notes.md",{"path":101,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":102,"sourceHtmlUrl":103},"references\u002Flayouts.md",13858,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Flayouts.md",{"path":105,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":106,"sourceHtmlUrl":107},"references\u002Fppt-layout-patterns.md",8003,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Fppt-layout-patterns.md",{"path":109,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":110,"sourceHtmlUrl":111},"references\u002Fstyles.md",27799,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Fstyles.md",{"path":113,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":114,"sourceHtmlUrl":115},"references\u002Ftroubleshooting.md",12401,"https:\u002F\u002Fgithub.com\u002Fsapsapshen\u002Fai-memory-router\u002Fblob\u002Fmain\u002Fskills-store\u002Fhermes\u002Fstable-diffusion\u002Freferences\u002Ftroubleshooting.md",{"path":117,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":118,"sourceHtmlUrl":119},"references\u002Fwechat-extraction.md",997,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Freferences\u002Fwechat-extraction.md",{"path":121,"resourceType":122,"resourceTypeDesc":123,"sizeBytes":124,"sourceHtmlUrl":125},"scripts\u002Fcheck_deck.py","SCRIPT","脚本",9778,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Fscripts\u002Fcheck_deck.py",{"path":127,"resourceType":122,"resourceTypeDesc":123,"sizeBytes":128,"sourceHtmlUrl":129},"scripts\u002Fcheck_plan.py",6229,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Fscripts\u002Fcheck_plan.py",{"path":131,"resourceType":122,"resourceTypeDesc":123,"sizeBytes":132,"sourceHtmlUrl":133},"scripts\u002Fpack_core.py",3024,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Fscripts\u002Fpack_core.py",{"path":135,"resourceType":122,"resourceTypeDesc":123,"sizeBytes":136,"sourceHtmlUrl":137},"scripts\u002Fprepare_source.py",7431,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Fscripts\u002Fprepare_source.py",{"path":139,"resourceType":76,"resourceTypeDesc":77,"sizeBytes":140,"sourceHtmlUrl":141},"template.html",30582,"https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fnote-slides\u002Ftemplate.html",[143,194,218,257,279,309],{"skillNo":144,"name":145,"nameZh":6,"description":146,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":147,"licenseFileUrl":148,"sourceHtmlUrl":149,"categories":150,"keywords":165,"publishedAt":38},"01M6DCCDBC0GW6XYQYM9Z0TRYZ","xlsx","Use this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like \"the xlsx in my downloads\") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.","Apache-2.0","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fxlsx\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fxlsx\u002FSKILL.md",[151,154,157,160],{"categoryNo":152,"categoryName":153,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000002","文档处理",{"categoryNo":155,"categoryName":156,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000004","电子表格",{"categoryNo":158,"categoryName":159,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":161,"categoryName":162,"taxonomyType":163,"taxonomyTypeDesc":164},"01JBSEEDCAT000000000000034","行政与办公人员","PROFESSION","适用职业",[145,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193],"do","downloads","google sheets","Word","spreadsheet","pipeline","time","trigger","python script","fix","html report","python","api integration","open","files","use","document","database","done","sources","standalone","new","when","api","create","references","Charting","data sources",{"skillNo":195,"name":196,"nameZh":6,"description":197,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":147,"licenseFileUrl":198,"sourceHtmlUrl":199,"categories":200,"keywords":202,"publishedAt":38},"01M6DCBGARPB3ZTJ6C46J4K7B4","pdf","Use this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text\u002Ftables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting\u002Fdecrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fpdf\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fpdf\u002FSKILL.md",[201],{"categoryNo":152,"categoryName":153,"taxonomyType":22,"taxonomyTypeDesc":23},[203,204,205,166,206,180,207,181,208,209,210,211,212,187,213,214,215,216,217],"PDF","images","ocr","pages","produce","file","text","forms","user","extracting","reading","tables","make","with","to do",{"skillNo":219,"name":220,"nameZh":6,"description":221,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":222,"licenseFileUrl":223,"sourceHtmlUrl":224,"categories":225,"keywords":232,"publishedAt":38},"01M6DCB0HDC3VCDSCY65R7Q9C1","guizang-social-card-skill","Generate Guizang-style social card image sets, Live Photo motion cards, material-first Live Photo puzzle layouts, triple Live Photo collages, long-video-to-Live-Photo treatments, and WeChat official account cover pairs from articles, scripts, screenshots, product notes, subtitles, photos, or user-supplied videos. Use when the user asks for 小红书图文, Rednote\u002FXiaohongshu images, social cards, carousel images, 3:4 covers, Live Photo, 实况照片, 单视频实况拼图, 二宫格实况拼图, 三连实况拼图, 四宫格实况拼图, 微信公众号封面, WeChat 21:9 + 1:1 covers, Swiss Style, or magazine-style social images.","AGPL-3.0","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fguizang-social-card-skill\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fguizang-social-card-skill\u002FSKILL.md",[226,229],{"categoryNo":227,"categoryName":228,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000016","图像与多媒体",{"categoryNo":230,"categoryName":231,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000023","Skill与智能体开发",[233,234,235,236,204,237,238,239,240,181,241,242,243,244,245,246,247,248,249,250,211,251,252,253,254,188,34,255,256],"social","guizang social card skill","card","guizang","Video","official","wechat","motion","subtitles","image","xiaohongshu","notes","first","product","generate","live","photos","视频","screenshots","rednote","carousel","微信","scripts","cover",{"skillNo":258,"name":259,"nameZh":6,"description":260,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":222,"licenseFileUrl":261,"sourceHtmlUrl":262,"categories":263,"keywords":266,"publishedAt":38},"01M6DCAV90EX528Q36VRJH4VN4","guizang-ppt-skill","生成横向翻页网页 PPT（单 HTML 文件），含 WebGL 背景、演讲者视图、观众屏同步、讲稿备注、章节幕封、数据大字报、图片网格等模板。提供两种风格：① \"电子杂志 × 电子墨水\"（衬线 + 流体背景 + 暖色） ② \"瑞士国际主义\"（无衬线 + 网格点阵 + IKB\u002F柠檬黄\u002F柠檬绿\u002F安全橙高亮）。当用户需要制作分享 \u002F 演讲 \u002F 发布会风格的网页 PPT，或提到\"杂志风 PPT\"、\"瑞士风 PPT\"、\"Swiss Style\"、\"horizontal swipe deck\"时使用。","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fguizang-ppt-skill\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fguizang-ppt-skill\u002FSKILL.md",[264,265],{"categoryNo":227,"categoryName":228,"taxonomyType":22,"taxonomyTypeDesc":23},{"categoryNo":230,"categoryName":231,"taxonomyType":22,"taxonomyTypeDesc":23},[267,31,236,268,269,270,271,272,242,32,273,35,274,275,276,277,278],"guizang ppt skill","guizang ppt","deck","webgl","生成","安全","演讲","swipe","ikb","swiss","style","数据",{"skillNo":280,"name":281,"nameZh":6,"description":282,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":147,"licenseFileUrl":283,"sourceHtmlUrl":284,"categories":285,"keywords":288,"publishedAt":38},"01M6DCA9W98X2GJ6E6PPNWATAF","docx","Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fdocx\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fdocx\u002FSKILL.md",[286,287],{"categoryNo":152,"categoryName":153,"taxonomyType":22,"taxonomyTypeDesc":23},{"categoryNo":161,"categoryName":162,"taxonomyType":163,"taxonomyTypeDesc":164},[281,204,289,166,169,290,291,292,180,207,181,182,293,208,294,190,295,296,297,298,212,299,300,188,301,302,303,304,305,306,307,308],"as","page","triggers","general","documents","tasks","memo","unrelated","generation","template","professional","word documents","general coding","read","edit","spreadsheets","google","google docs","doc","coding",{"skillNo":310,"name":311,"nameZh":6,"description":312,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":313,"sourceHtmlUrl":314,"categories":315,"keywords":323,"publishedAt":38},"01M6DCA42A21D13JTWE4DNKHWS","darwin-skill","Darwin Skill 2.0 (达尔文.skill 2.0): autonomous skill optimizer, v2.0 integrates Microsoft Research SkillLens (arXiv 2605.23899) 9-dim rubric + SkillOpt (arXiv 2605.23904) validation-gated design + human-in-the-loop checkpoints. Evaluates SKILL.md files using a 9-dimension rubric (structure + effectiveness + meta-skill blacklists), runs hill-climbing with git version control, spawns independent judge agents for blind evaluation, validates improvements through test prompts with auto-break on diminishing returns, and generates visual result cards. Use when user mentions \"优化skill\", \"skill评分\", \"自动优化\", \"auto optimize\", \"skill质量检查\", \"达尔文\", \"darwin\", \"帮我改改skill\", \"skill怎么样\", \"提升skill质量\", \"skill review\", \"skill打分\".","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fdarwin-skill\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fiasiv5\u002Fskills\u002Fblob\u002Fmain\u002Fdarwin-skill\u002FSKILL.md",[316,319,322],{"categoryNo":317,"categoryName":318,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":320,"categoryName":321,"taxonomyType":22,"taxonomyTypeDesc":23},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":230,"categoryName":231,"taxonomyType":22,"taxonomyTypeDesc":23},[324,325,326,327,328,329,330,331,180,332,333,181,334,335,336,337,338,339,340,341,342,343,188,344,345,346,347],"darwin skill","darwin","dim","SkillOpt","meta-skill","human-in-the-loop","Meta","design","autonomous","auto","loop","version","visual","optimizer","review","test","structure","optimize","research","skill 2","human","result","break","Evaluation"]