把人与 AI 协同进化中跑通的工作流，整理成可以直接复用的能力包。
Reusable capability packs for human-AI co-evolution in real work.
一套可以直接下载、安装和使用的中文 AI Skills。 如果某个 Skill 帮你节省了时间，请点右上角 ⭐ Star 收藏这个仓库。你的 Star，也决定我下一步优先更新哪个 Skill。
很多人用 AI，问题不在于不会写提示词。
真正的问题是：每次都从零开始。
今天让 AI 翻译一份 PDF，明天让 AI 整理一篇文章，后天又想做一个可以分享的读物。每次都要重新解释背景、重新讲标准、重新调格式。时间久了，你会发现自己不是在用 AI 提效，而是在不断训练一个临时工。
这个仓库想解决的是另一件事：
把人与 AI 一起做事、一起学习、一起改进时跑通的工作流，沉淀成一个个可以复用、可以安装、可以分享的「能力包」。
它不是一个 prompt 大杂烩，也不是我的私人配置备份。
它更像一个逐步长出来的工具箱：每个能力包都应该能完成一个真实任务，并且有清楚的输入、输出、边界和使用方式。
本仓库包含罗老师原创并持续维护的 AI Skills，也会收录经过实际测试的社区优秀 Skills。社区 Skill 的版权归原作者所有，我们优先链接原仓库，并鼓励用户支持原作者。
| 标识 | 类型 | 仓库处理方式 |
|---|---|---|
🔵 LLS Original |
罗老师原创 | 收录完整源码和安装包 |
🟡 LLS Adapted |
在许可证允许范围内改编 | 标明原作者、原仓库、许可证和修改内容 |
🟢 Community Pick |
罗老师核验推荐 | 默认只提供中文介绍和原仓库链接 |
社区 Skill 由原作者创作，请优先前往原仓库给作者一个 ⭐ Star。 如果你认可罗老师的筛选、测试和中文说明，也欢迎 ⭐ Star 本导航仓库。
覆盖 Skill 创建、标题、命名、Mermaid、截图隐私、最佳实践、GitHub 首页、决策辩论、文件整理、纸质文件修复、视频学习、安全审计、中文幽默、真人化写作、短视频拆解和国际技术文档。完整清单见 registry.json。
另有 63 个 Community Pick 仅做中文介绍和上游导航。第三方推荐不复制源码和安装包，也不提交 SkillHub。
女娲、达尔文和同事 Skill 都是罗老师推荐的优秀社区项目。本仓库只解释它们分别适合什么场景，并把用户引导到原作者仓库，不复制源码、不重新包装成罗老师改编版。
| 项目 | 适合做什么 | 支持原作者 |
|---|---|---|
dot-skill（同事.skill） |
蒸馏“人怎么做事” | 打开原始仓库 |
nuwa-skill（女娲.skill） |
蒸馏“高手怎么思考” | 打开原始仓库 |
darwin-skill（达尔文.skill） |
测试并进化已有 Skill | 打开原始仓库 |
详细中文说明见 Community Pick。喜欢具体 Skill，请优先给原作者 Star。
| 能力包 | 来源 | 它解决什么问题 | 状态 |
|---|---|---|---|
lls-word-memory-coach |
🔵 LLS Original | 用翻译、拼读、可信构词与联想记忆单词，再做回忆小测。 | Beta · 源码可用 |
bilingual-reader-maker |
🔵 LLS Original | 把英文 PDF / 长文做成中英双语移动阅读版，输出 HTML + PDF。 | Stable |
concept-explainer |
🔵 LLS Original | 在做真实任务时，把陌生工具、命令、配置、架构和报错链路解释成通俗知识卡片。 | Stable |
lls-workbuddy-guide |
🔵 LLS Original | 带新手从真实任务开始，跑通 WorkBuddy 的执行、验收与复用。 | Stable |
lls-ppt-briefing-coach |
🔵 LLS Original | 用七阶段访谈澄清 PPT 需求，生成可执行的制作指令。 | Stable |
lls-learning-guide |
🔵 LLS Original | 陪孩子逐步答题、费曼复述并沉淀学习档案。 | Stable |
lls-skill-lifecycle-manager |
🔵 LLS Original | 隔离管理私有母库与公开分享库的版本、发布和运行副本。 | Stable |
lls-dialogue-subtext-reader |
🔵 LLS Original | 分开原话事实与推测，生成可验证、低冲突的沟通回应。 | Stable |
lls-local-excel-vba-data-processor |
🔵 LLS Original | 用本地 Excel VBA 处理敏感数据，附测试、日志、回滚和验收。 | Stable |
lls-ip-positioning-coach |
🔵 LLS Original | 用七步互动找到真实可信、能持续表达并可小步验证的个人 IP 定位。 | Stable |
lls-mermaid-diagram |
🔵 LLS Original | 选择正确图型、压缩结构并真实编译 Mermaid 图。 | Stable |
lls-ppt-workflow-builder |
🔵 LLS Original | 从演示目标推进到逐页制作、演练与验收。 | Stable |
lls-prompt-basics-coach |
🔵 LLS Original | 用真实任务和样例测试教普通用户写提示词。 | Stable |
lls-code-anatomy |
🔵 LLS Original | 把一小段代码拆成数据流、语法解剖、亲手运行和可复习学习卡片。 | Stable |
lls-short-video-breakdown |
🔵 LLS Original | 用时间点和原文证据拆解短视频结构、留存、口语与镜头。 | Stable |
lls-douyin-oral-script-writer |
🔵 LLS Original | 把真实观点写成自然、可信并经过朗读计时的抖音口播稿。 | Stable |
lls-ai-video-script-planner |
🔵 LLS Original | 从传播任务推进到分镜、素材、生成提示和制作验收。 | Stable |
lls-ai-drawing-prompt-coach |
🔵 LLS Original | 从画面意图到可测试、可迭代的文生图提示词。 | Stable |
lls-ai-poster-design-coach |
🔵 LLS Original | 从传播目标和信息层级推进到海报版式与导出验收。 | Stable |
lls-jimeng-image-editing-coach |
🔵 LLS Original | 保留原图关键结构，分区、小步完成图片编辑。 | Stable |
lls-ai-data-analysis-starter |
🔵 LLS Original | 从业务问题和数据质量开始，完成可复核的数据分析。 | Stable |
lls-government-publicity-data-analyst |
🔵 LLS Original | 用统一口径分析政务传播与公共服务效果。 | Stable |
lls-geo-brand-visibility-coach |
🔵 LLS Original | 用问题集、事实库和来源审计管理生成式搜索可见性。 | Stable |
lls-aida-marketing-copywriter |
🔵 LLS Original | 以真实证据和渠道约束写 AIDA 营销文案。 | Stable |
lls-leader-speechwriter-expert |
🔵 LLS Original | 基于身份、场合和真实材料写正式讲话稿。 | Stable |
lls-slide-storyline-planner |
🔵 LLS Original | 从听众认知变化到章节逻辑和逐页结论标题。 | Stable |
lls-ai-microdrama-planner |
🔵 LLS Original | 从冲突发动机推进到逐集节拍、悬念账本和 AI 制作连续性。 | Stable |
lls-ai-novel-prompt-engineer |
🔵 LLS Original | 用分层上下文、场景循环和写后回填管理长篇小说。 | Stable |
lls-education-illustration-maker |
🔵 LLS Original | 从学习目标和准确性约束制作可验证的教育插图。 | Stable |
lls-article-screenshot |
🔵 LLS Original | 从证据地图推进到最小范围截图、规范命名和隐私复核。 | Stable |
lls-document-print-restorer |
🔵 LLS Original | 保守增强可见内容，生成 A4 PDF、预览和哈希核对报告。 | Stable |
lls-image-privacy-redactor |
🔵 LLS Original | 按确认坐标生成脱敏副本、哈希和整图复核报告。 | Stable |
lls-github-profile-writer |
🔵 LLS Original | 从访客任务和真实资产设计 GitHub 首屏、代表作和快速开始。 | Stable |
lls-headline-summary-maker |
🔵 LLS Original | 从正文主张生成多机制标题、摘要并检查兑现度。 | Stable |
lls-human-writing-editor |
🔵 LLS Original | 锁定事实和作者声音，分层去除模板腔。 | Stable |
lls-chinese-humor |
🔵 LLS Original | 按受众、场合和边界设计可朗读的中文幽默。 | Stable |
lls-mirror-debate |
🔵 LLS Original | 用钢人化论证、证据质询和可逆试验压力测试决策。 | Stable |
lls-naming-coach |
🔵 LLS Original | 从命名约束到候选策略、场景测试和正式核验清单。 | Stable |
第一个能力包来自一个很具体的需求：
把一份英文创业手册，做成适合中文读者学习英语的双语读物。
最后我们定下来的标准是：
- 保留原材料的视觉气质，比如封面、章节图、配色、图标。
- 正文不强行左右分栏，而是手机友好的单栏阅读。
- 英文一段，中文一段，方便查词、跟读、复制和解析。
- HTML 是真文本，不是截图；PDF 从 HTML 导出，方便分发。
- 署名放在文末，克制一点，不打扰阅读。
因为对普通使用者来说，skill 这个词还是太技术。
他们真正关心的不是：
这个 prompt 怎么写？
这个 agent 怎么调用？
这个脚本怎么跑？
他们关心的是：
我给你一个英文 PDF，你能不能变成一份好看的双语读物？
我有一批笔记，你能不能帮我整理成文章？
我有一个工作流，能不能下次不用重新解释？
所以这里的每个 skill，都会尽量被打磨成一个「能力包」：
让 AI 不只是聊天，而是稳定地完成一类任务。
原创 Skill 的 SKILL.md、references、README/docs 和原创素材采用 CC BY-NC-SA 4.0；skills/**/scripts/** 与 tools/** 中的实质软件代码采用 PolyForm Noncommercial 1.0.0。adapted/、community/ 与第三方材料保持各自上游许可证。
企业内部普通办公可按额外许可免费使用。收费课程、转售、客户交付、SaaS、代运营及其他商业使用须取得商业授权。历史 MIT 版本已经授予的权利继续有效；完整分流说明见 LICENSE 和 第三方通知。
每个 Skill 页面都写明唯一 slug，并提供两条安装路径。最省事的方式是在 WorkBuddy 对话中粘贴：
请按 https://skillhub.cn/install/skillhub.md 检查 SkillHub CLI。
搜索目标 Skill 的精确 slug；只有 slug 完全一致时，才安装到 `~/.workbuddy/skills/`。
安装后读取 `~/.workbuddy/skills/<slug>/SKILL.md`，核对 name 和 version（如有），再新开或重启 WorkBuddy 会话。
也可在左侧 「技能」 → 「添加技能」或「查找技能」 中搜索精确 slug 后安装；不同 WorkBuddy 版本的界面名称可能略有变化。若没有精确匹配，请不要安装同名条目，改用该 Skill README 的 GitHub Release 或手动复制目录。
# Codex
mkdir -p ~/.codex/skills
cp -R skills/bilingual-reader-maker ~/.codex/skills/
# Claude Code
mkdir -p ~/.claude/skills
cp -R skills/bilingual-reader-maker ~/.claude/skills/安装后开启新的会话，让客户端刷新 Skill 列表。
私有母库与公开分享库各自是真源
两库之间没有复制、镜像、同步或一键公开
GitHub 只承载公开分享库的源码与下载
手机/H5 是普通人的入口
Skill/CLI 是进阶用户的工具层
这个仓库会优先保证三件事：
- 可复用：不是一次性的聊天记录，而是能重复执行的流程。
- 可理解：尽量讲人话，让非技术用户也知道它能干什么。
- 可公开：不放私人路径、账号密钥、个人数据库、微信文件路径和不可分享的材料。
这个仓库会慢慢加入更多能力包：
- 双语读物制作
- 任务中的概念解释与知识卡片
- 文章与公众号发布准备
- 知识库整理
- 学习材料转 HTML / PDF
- 面向普通人的 AI 工作流模板
如果你只是想看热闹，可以先 Star。
如果你也在把自己的 AI 工作流产品化，欢迎 fork、改造，或者提 issue 分享你的场景。
- ⭐ Star：收藏项目、表达支持，方便以后找回来。
- 🔔 Watch → Custom → Releases：接收新版本发布通知。
- 📦 Download：在每个 Skill 页面下载对应的 ZIP 安装包。
如果某个 Skill 帮你节省了时间，请点右上角 ⭐ Star 收藏这个仓库。你的 Star，也决定我下一步优先更新哪个 Skill。
Most people do not fail with AI because they lack prompts.
They fail because every workflow starts from scratch.
One day you ask AI to translate a PDF. The next day you ask it to turn an article into a shareable document. Then you need a mobile-friendly reader, a publishing checklist, or a reusable writing process. Each time, you explain the background again, restate the standard again, and fix the format again.
This repository is built around a different idea:
Turn proven human-AI co-evolution workflows into reusable capability packs.
It is not a random prompt dump.
It is not a private config backup.
It is a growing library of practical skills that are meant to do real work.
- 🔵 LLS Original: created and maintained by Luo Laoshi.
- 🟡 LLS Adapted: transparently adapted under the upstream license.
- 🟢 Community Pick: curated, provenance-checked recommendation linking to the original repository by default.
Community skills remain the property of their original authors. Please support the upstream author first; Star this repository as well if the testing and Chinese guidance help you.
The Community Pick directory currently contains 63 link-only recommendations. Third-party source code and packages are not republished, and these entries are not submitted to SkillHub.
| Capability Pack | What it does | Status |
|---|---|---|
bilingual-reader-maker |
Turns English PDFs or long articles into polished Chinese-English mobile readers, with HTML and PDF output. | Stable |
concept-explainer |
Explains unfamiliar tools, commands, configs, architecture, and troubleshooting chains as short knowledge cards during real tasks. | Stable |
lls-workbuddy-guide |
Guides beginners through real WorkBuddy tasks, verification, and reusable workflows. | Stable |
lls-ppt-briefing-coach |
Clarifies presentation needs through a seven-stage interview and produces an actionable brief. | Stable |
lls-learning-guide |
Guides students through problem solving, Feynman retelling, and learning records. | Stable |
lls-skill-lifecycle-manager |
Governs the full skill lifecycle across source, releases, documentation, and runtime copies. | Stable |
lls-dialogue-subtext-reader |
Separates observable dialogue facts from interpretations and produces verifiable, low-conflict replies. | Stable |
lls-local-excel-vba-data-processor |
Builds privacy-preserving local Excel VBA workflows with tests, logs, rollback, and acceptance checks. | Stable |
lls-ip-positioning-coach |
Uses a seven-step coaching flow to create credible, sustainable, testable personal-brand positioning. | Stable |
lls-mermaid-diagram |
Selects the right diagram type and validates Mermaid by real compilation. | Stable |
lls-ppt-workflow-builder |
Runs a presentation from decision goal and evidence through production and rehearsal. | Stable |
lls-prompt-basics-coach |
Teaches prompting through clear goals, constraints, formats, and test cases. | Stable |
lls-short-video-breakdown |
Breaks down short videos with timestamps, evidence, retention logic, delivery, and visuals. | Stable |
lls-douyin-oral-script-writer |
Turns real ideas and evidence into natural, timed Douyin spoken scripts. | Stable |
lls-ai-video-script-planner |
Plans video beats, storyboards, assets, generation prompts, and production acceptance. | Stable |
lls-ai-drawing-prompt-coach |
Turns visual intent into testable and iterative text-to-image prompts. | Stable |
lls-ai-poster-design-coach |
Designs poster information hierarchy, layout, assets, and export acceptance. | Stable |
lls-jimeng-image-editing-coach |
Preserves key image structure through masked, single-variable editing and inspection. | Stable |
lls-ai-data-analysis-starter |
Runs reproducible analysis from business questions and data quality checks. | Stable |
lls-government-publicity-data-analyst |
Measures public communication and service outcomes with governed metrics. | Stable |
lls-geo-brand-visibility-coach |
Manages generative-search visibility through question sets, facts, sources, and retesting. | Stable |
lls-aida-marketing-copywriter |
Writes evidence-led AIDA marketing copy under channel and truthfulness constraints. | Stable |
lls-leader-speechwriter-expert |
Produces formal speeches from verified material, role, occasion, timing, and review. | Stable |
lls-slide-storyline-planner |
Plans audience change, core claim, evidence, chapters, and slide headlines. | Stable |
The first pack came from a concrete use case: turning an English startup playbook into a bilingual study edition for Chinese readers.
The resulting standard is simple:
- Keep the source material's visual identity when appropriate.
- Do not force side-by-side columns when they hurt mobile reading.
- Put the English paragraph first, then the Chinese translation.
- Keep HTML as real text, so it can be copied, searched, parsed, and restyled.
- Export PDF from HTML for easy sharing.
Because most users do not care about the internal words: prompt, agent, script, tool call.
They care about the job:
Can you turn this PDF into a good bilingual reader?
Can you convert my notes into a publishable article?
Can we make this workflow reusable next time?
That is what each skill here is trying to become:
a reusable unit of work that makes AI more dependable.
Each skill README contains its exact slug and two WorkBuddy installation paths. In a WorkBuddy chat, ask it to follow https://skillhub.cn/install/skillhub.md, search the exact slug, install only an exact match to ~/.workbuddy/skills/, then read back ~/.workbuddy/skills/<slug>/SKILL.md and start a new session. The left Skills → Add/Find Skill flow is the equivalent UI path; labels may vary by version.
If no exact slug is found, do not install a similarly named skill. Use the release or manual-copy fallback documented for that skill instead. See the full WorkBuddy install guide.
# Codex
mkdir -p ~/.codex/skills
cp -R skills/bilingual-reader-maker ~/.codex/skills/
# Claude Code
mkdir -p ~/.claude/skills
cp -R skills/bilingual-reader-maker ~/.claude/skills/Start a new session after installation.
The private factory and public sharing repository are separate sources of truth.
There is no copy, mirror, sync, or one-click-public path between them.
GitHub serves only the public sharing repository.
Mobile/H5 is the beginner-friendly front door.
Skills and CLIs are the advanced working layer.
This repository will grow slowly, with a bias toward useful, understandable, and shareable workflows.
- ⭐ Star saves the project and shows your support.
- 🔔 Watch → Custom → Releases follows new release notifications.
- 📦 Download gets the installable ZIP from each skill page.
Original Skill content (SKILL.md, references, README/docs, and original assets) is licensed under CC BY-NC-SA 4.0. Material software code in skills/**/scripts/** and tools/** is licensed under PolyForm Noncommercial 1.0.0. Adapted, community, and third-party material remains under its upstream license.
Free additional permission is available for ordinary internal office use; see ADDITIONAL-PERMISSIONS.md. Paid courses, resale, client delivery, SaaS, managed operations, and other commercial uses require a commercial license. Rights granted for historical MIT releases continue; see LICENSE and THIRD_PARTY_NOTICES.md.