WorkBuddy 专家团 Skill 包（开源版）

使用提示：本仓库所含 skill 文件均由用户在 WorkBuddy 平台官方渠道、通过自己个人账号、以普通对话方式生成。未使用破解安装包 / 获取应用文件夹内文件 / 反编译 / 抓包 / 模型 API 调用数据逆向等技术手段。不保证与 WorkBuddy 原版 100% 一致（平台侧可能更新，生成结果受 prompt / 模型版本影响）。

---

1. 项目简介

本仓库存档 WorkBuddy 平台（一个 AI 协作专家 agent 市场）的团队协作型 skill —— 每个 folder 对应一个"专家团"（多角色协作的 agent 集合），包含一个主理人 + 多个专业角色，工作流化地完成特定场景任务（投资研究 / 视频制作 / MVP 开发 / 财税 / 法务 / 营销 / 设计 / 法律 / 数据分析 等）。

Skill 内容主要由两类来源组装而成：

• WorkBuddy 官方团队（在 WorkBuddy 账号标记为 "WorkBuddy 团队"）
• 特邀作者 / 合作方（Easychen / 卡尔的AI沃茨 / 花叔 (Alchain) / 苍何 / 杜哥学量化 / 百望股份 / vincentlli / Excellent / AI科学局 / Xu Qingchu 等）

每个 skill 在导入 orchestration 框架（OpenAI Custom GPTs / Claude Project / Coze / Dify / MetaGPT / AutoGen / CrewAI 等）时相当于一份长 system prompt，可以整体调度主理人 → 使用子 agent 调用成员，完成复杂的多阶段协作。

---

2. 授权 / 免责声明

2.1 授权形式

代码 / 文档部分采用 MIT 许可证（详见 LICENSE 文件）。

2.2 数据来源与合规声明

1. 生成方式透明。所有 skill 文件均由用户在 WorkBuddy 平台官方渠道、通过自己的个人账号、以普通对话方式生成。未使用破解安装包、未获取应用文件夹内文件、未反编译、未抓包、未基于模型 API 调用数据逆向等任何技术手段。
2. 不保证一致性。skill 内容在不同时间 / 不同 prompt / 不同模型版本下可能会有细微差异，因此不保证与 WorkBuddy 原版 100% 一致。如你依赖其中任何信息（法条、财税规则、投资建议、医疗/安全关键内容），请务必自行核实。
3. 数据与合规。生成素材来自：合法的 WorkBuddy 平台调用输出、领域合规文档、公开语料。用户对自己在本仓下载 / 使用的内容负全部责任。
4. 第三方 attribution。部分 skill 直接复用了下列开源 / 特邀作者的方法论或文档，版权与归属仍属原作者：
   • 一人企业方法论（One Person Company）：Easychen — https://github.com/easychen/opc-methodology
   • AST-based Humanize PPT（MIT License）：LearnPrompt 团队 — https://github.com/LearnPrompt/humanize-ppt
   • 腾讯自选股股票投研系列：作者 jensonli@tencent.com
   • "刺桐说 Pro" 投资社群系列：杜哥学量化
   • "卡尔的人感 PPT" 系列：卡尔的AI沃茨
   • "AI 视频创作 / 视频解剖" 系列：苍何
   • "花叔数据分析" 系列：花叔 (Alchain)
   • "智能发票" 系列：百望股份
   • "Makers 开发" 系列：vincentlli
   • "MVP 开发" 系列：Excellent
   • "NCRE 等级考试" 系列：Xu Qingchu
   • "KET 备考" 系列：AI科学局

如你为原始作者且希望修改或移除相关内容，请开 Issue — 我们会在 5 个工作日内响应。

---

3. 分类索引

📖 阅读建议：找到你关心的类别 → 点开对应目录 → 优先看 README.md（没有 README 的以 SKILL.md 或主理人 *-team-lead.md 为准）。 一个 skill 可能出现在多个类别（如 seo-content-skills 同时属于"内容营销"与"SEO"），下表按主类别只列一次，在"复用视角"段指出跨域协作。

3.1 投资分析 📈 (6 个专家团 / ~50+ 位专家)

新目录	简介	专家数（主理人 + 成员）
investment-masters-skills	AI 式对冲基金圆桌 — 13 位传奇投资哲学家 + 6 位分析师 + 2 位管理人（投资组合经理 + 风控师），参考巴菲特 / 芒格 / 费雪 / 邓普顿 / 达利欧 / ・拉什・库马尔等	21
a-share-skills	A 股全链路研究：宏观策略 → 产业映射 → 资金追踪 → 个股深度 → 估值 → 风险诊断 → 投顾	8
stock-partner-skills	腾讯自选股实战派：产业策略 / 信号捕捉 / 估值 / 抄底 / 基本面 / 短线，6 大真实炒股视角	7
trading-agent-skills	多空辩论式投研：技术面 / 基本面 / 新闻 / 情绪 / 多空研究员 / 研究总监 / 交易员 / 三方风险	13
citongshuopro-skills	刺桐说 Pro 投资社群：主理人调度 4 位社群嘉宾（Gy / 贾总 / 小星 / 张老师）	4
super-partner-skills	福帮手超级合伙人：入口路由 → 公司下一步 / 天才合伙人 / 多维度协同工作坊 / 服务跟进 / 乐包商业化	11

3.2 变现 / 商业化 💰 (专家团复用)

新目录	复用场景
content-monetization-skills	CPS 带货 / CPE·CPM 效果广告 / 创作者 ↔ 品牌交易撮合 / BD — 5 人商业化闭环
super-partner-skills	商业化落地：乐包激活与能力继续（Lebao），把能力调用转成商业计划
a-share-skills / stock-partner-skills	A 股 / 投研内容→ 个人 IP 素材
investment-masters-skills	大师方法论（dhandho-master 等）→ 付费课程 / 知识产品

3.3 内容创作 / 营销 📣 (10 个专家团 / ~55+ 位专家)

新目录	简介	专家数
ai-content-creator-skills	多模态内容生产：创意策略 / 文案 / 影像生成 / 精修合成 / 素材改编 / AI 视频	7+
video-gen-skills	AI 短视频团队：采集（灵阅）→ 策划脚本（灵枢）→ 渲染 MP4（灵映）+ 引用素材库	5
promo-creator-skills	产品宣传片：创意简报 → 逐镜头分镜 → 素材生产 → HyperFrames 剪辑 → BGM → 交付	6
content-distribution-skills	跨 13+ 平台多平台分发（含微信视频号、小红书）	5
marketing-campaign-skills	营销战役：内容创作 / 活动策划 / SEO / 品牌分析 — 4 人全生命周期	4
seo-content-skills	SEO 内容营销：关键词研究 / 技术优化 / 长文创作 / 内容编辑 / 链接策略 / 转化率分析	5+
ket-prep-skills	KET 剑桥少儿备考：学情诊断 / 词汇语法 / 听读 / 写说 / 模考	5
ncre-exam-skills	NCRE 计算机等级：一级→四级 + 主理人（徐庆初）	5
open-spec-doc-skills	企业级长文档：调研 / 生成 / 审核 — 6 阶段 SOP	3
humanize-ppt-skills	AST 人感 PPT：大纲导演 → 魁藏 / Zara / HyperFrames / OPC / 前端 HTML / 视频动效	5+

3.4 内容细分 — 教学 / 科研 / 智能 🎓 (4 个专家团)

新目录	简介	专家数
academic-journal-selector-skills	学术选刊顾问 v3.0：中外刊并行管道，冲稳保三层投稿方案	4
gpt-researcher-skills	深度研究：5 阶段 SOP，多源聚合 + 审稿循环	6
ket-prep-skills	同上（语言教学复用）	—
ncre-exam-skills	同上（考试复用）	—

3.5 开发 / 工程 DevOps 🛠 (7 个专家团 / ~40+ 位专家)

新目录	简介	专家数
aicoding-architecture-expert-skills	复杂系统架构：摄入→调研→高层→系统→UserStory→部署→安全 + Gate 返工	6
mvp-dev-skills	MVP 全栈：PM / 设计 / 架构 / 前端 / 后端 / QA / DevOps	7
makers-skills	EdgeOne Makers 部署：前端 / 后端 / AI Agent / CLI / 部署 + 交付总监	6
software-company-skills	软件公司（快速模式）：PM 定需求 → 架构师拆任务 → 工程师批量实现 → QA	5
software-workshop	GStack 软件工坊：产品评审 / 代码审查 / 安全审计 / QA / 设计系统 / 调试运维 + 主理人	6
engineering-assurance-skills	工程保障：架构评审 / 代码审查 / SRE 事故响应 / 测试 / 技术文档	5
ai-data-copilot-skills	智数分析 (Crew AI)：SQL / 数据科学 / RAG / 可视化 / 报告生成	6

3.6 设计 Design 🎨 (2 个专家团 / ~11 位专家)

新目录	简介	专家数
design-engine-skills	设计原型：发现 / 设计系统 / 原型 / 评审 / 导出 + 主理人 + 内置 71 套设计系统	6
humanize-ppt-skills	（见 3.3）AST 方法论上游，作为设计工程的一支	—

3.7 法律 / 合规 Legal ⚖ (3 个专家团 / ~15 位专家)

新目录	简介	专家数
chatlaw-skills	中文法律咨询：民事 / 婚姻 / 合同 / 劳动 / 侵权 — 4 阶段 + 判例撰写	5
enterprise-legal-skills	企业法务：合同 / 并购 / 劳动 / 隐私 / 产品 / 合规 / AI 治理 / IP + 中国区本地化	9
smb-team-skills	SMB 经营总管：客户 / 合约 / 投诉 / CRM / 流失预警	1

3.8 财税 Tax / Compliance 🧾 (2 个专家团 / ~12 位专家)

新目录	简介	专家数
tax-compliance-skills	财税合规：票据 / 记账 / 报表 / 税务申报 / 合规审计 + 主理人 + 引擎	6
invoice-verify-skills	智能发票（百望体系）：识别 → 验真 → 征信 → 归档	5

3.9 数据分析 Data Analysis 📊 (2 个专家团 / ~10 位专家)

新目录	简介	专家数
huashu-data-pro-skills	花叔本地数据分析：趋势 / 结构 / 异常三专家并行，网页 / Excel / PPT 三格式报告	3+
ai-data-copilot-skills	智数分析（应用化 SQL 工作流，侧重应用系统落地）	—

3.10 销售 Sales 🤝 (1 个专家团 / 5 位专家)

新目录	简介	专家数
sales-battle-skills	销售攻坚：客户研究 / 竞品情报 / 外联策略 / 销售预测 — 主理人 + 4 成员	5

3.11 运营 / HR Ops 🏢 (1 个专家团 / 5 位专家)

新目录	简介	专家数
hr-operations-skills	HR 运营：招聘 / 薪酬分析 / 组织发展 / 入职管理 + 主理人	5

3.12 社媒互动 Social Growth 📱 (1 个专家团 / 5 位专家)

新目录	简介	专家数
social-engagement-skills	社媒增长：AI 评论 / 互动自动化 / 品牌舆情 / 信号挖掘 — 覆盖 14+ 平台	5

3.13 市场 / 产品 📈 (2 个专家团 / ~9 位专家)

新目录	简介	专家数
product-strategy-skills	产品战略：需求 / 用户研究 / 竞品 / 数据 / 路线图 + 主理人	5
gpt-researcher-skills	（见 3.4，深度研究复用）	—

---

4. 快速开始

4.1 导入到任意 agent / 编排框架

每个 SKILL.md 本质上是一段长 system prompt。典型用法：

• 直接用作 OpenAI Custom GPTs / Claude Project / Gemini / Coze / Dify / MetaGPT / AutoGen / CrewAI 的 system prompt
• 把"知识检索工具"调用改成目标平台的等价工具即可跨框架移植
• 最优结构：以 *team-lead/SKILL.md 作为 orchestrator，成员 skill 按需作子 agent 调用

4.2 目录命名规则

本仓统一改为 kebab-case-skills 风格：

原 WorkBuddy 团队名	仓内目录
InvestmentGroups	investment-masters-skills
TradingAgentTeam	trading-agent-skills
AiContentCreatorTeam	ai-content-creator-skills
AiDataCopilot	ai-data-copilot-skills
DesignEngineTeam	design-engine-skills
刺桐说Pro	citongshuopro-skills
学术选刊顾问团	academic-journal-selector-skills
社媒互动增长专家团	social-engagement-skills
花叔数据分析专家团	huashu-data-pro-skills
StockPartnerTeam	stock-partner-skills
GPTResearcherTeam	gpt-researcher-skills
SeoContentTeam	seo-content-skills
MvpDevExpertTeam	mvp-dev-skills
…	…

完整对照表请查看附录 A。

4.3 跨 skill 协作参考

本仓 skill 不强制依赖其他目录，但一些 skill 在 description: 段提到协作场景：

• mvp-dev-skills ↔ makers-skills ↔ aicoding-architecture-expert-skills：前端 / 后端 / Agent 提示注入的工程互参
• marketing-campaign-skills ↔ seo-content-skills：共享关键词策略 / 内容日历模板
• product-strategy-skills 下游可接到 ai-data-copilot-skills / huashu-data-pro-skills
• chatlaw-skills ↔ enterprise-legal-skills ↔ smb-team-skills：消费者 ↔ 企业 ↔ SMB 三层法律阶梯
• investment-masters-skills / a-share-skills / trading-agent-skills 三套投研互相 roundtable

4.4 推荐起始组合

目标	选用
营销 + 内容	ai-content-creator-skills + content-distribution-skills + marketing-campaign-skills
A 股投研	a-share-skills + stock-partner-skills + trading-agent-skills
大师圆桌	investment-masters-skills
MVP / App 交付	mvp-dev-skills 或 mvp-dev-skills + software-company-skills
设计 → 工程闭环	design-engine-skills + software-company-skills + software-workshop
数据 / 分析策略	ai-data-copilot-skills + huashu-data-pro-skills
内容变现	content-monetization-skills + promo-creator-skills
学术 / 教育 / 选刊	academic-journal-selector-skills + gpt-researcher-skills + ket-prep-skills + ncre-exam-skills
SMB 经营治理	enterprise-legal-skills + tax-compliance-skills + invoice-verify-skills + hr-operations-skills + smb-team-skills
投研 IP 变现	stock-partner-skills + investment-masters-skills + citongshuopro-skills

---

5. 参与贡献

欢迎任何能提高 skill 复用性 / 团队互操作性 / 可读性 / 可移植性 的贡献。

• 事实修复：若 SKILL.md 内容已与目录不一致 → 开 PR / Issue
• 翻译：任何尚未有 README.en.md 的 skill → 欢迎补英译
• SOP 模板化：把跨 skill 的 lead 检查单 / member 返工契约 / Gate 清单提出 _templates/
• 框架迁移：把 WorkBuddy 原生工具调用移植到 LangChain / CrewAI / AutoGen / Dify，放在 examples/
• 保留原作者 attribution：衍生 README 顶部保留原始作者 ID / 链接 / name: 字段

提交约定

1. PR 正文列出：涉及哪些目录、新增哪些模板、移除了哪些硬编码工具调用。
2. 禁止提交商业账号 id / token、API key 等敏感信息。
3. 如补 README.en.md，请与 README.zh.md 的章节号一致。

---

附录 A：旧 → 新目录名对照表

旧目录名（迁移前）	新目录名
InvestmentGroups	investment-masters-skills
SEOskills	seo-content-skills
aicoding-expert-team	aicoding-architecture-expert-skills
content-distribution-team	content-distribution-skills
copilot-skills	ai-data-copilot-skills
doc-team-skills	open-spec-doc-skills
engineering-assurance	engineering-assurance-skills
hr-operations-skill-files	hr-operations-skills
invoice-verify-skill-docs	invoice-verify-skills
ket	ket-prep-skills
marketing-campaign	marketing-campaign-skills
ncre-skills	ncre-exam-skills
opc-team-docs	opc-team-skills
product-strategy	product-strategy-skills
research-team-skills	gpt-researcher-skills
sales-battle	sales-battle-skills
smb-skills	smb-team-skills
stock-partner-skill-docs	stock-partner-skills
video-gen-team-skills	video-gen-skills
刺桐说Pro	citongshuopro-skills
学术选刊顾问团	academic-journal-selector-skills
社媒互动增长专家团	social-engagement-skills
花叔数据分析专家团	huashu-data-pro-skills

附录 B：仓库规模

• 39 个 skill 目录
• 450+ 个 md 文档（SKILL.md / README.md / agents / members / references）
• 13 个一级分类、~170+ 个独立专家角色
• 总仓库大小：~600KB 纯文本（平台描述性 skill 包，无模型权重）

---

6. 普通用户（非开发者）如何利用这些 expert skills

你不需要写代码，也能把这份 repo 里的许多 expert skills 立刻用在日常对话 AI 上。 核心思路：任何一个 SKILL.md 本质都是一段很长很详细的「系统提示词」；把它们粘贴到对话中，就相当于「召唤了那个专家」。

6.1 在国产免费 AI 中直接用（国内用户优先）

App	How to use	Good for
Doubao (豆包, ByteDance)	① Go to "智能体 / AI 应用" → "创建智能体" ② Paste a SKILL.md into the "系统提示词" (system prompt) ③ Keep trigger words as sample prompts in instructions	Copywriting / marketing / content / office / education
DeepSeek (深度求索)	① Paste the full SKILL.md at the start of the conversation or as a 【系统消息】 ② Multi-turn conversations will adhere to the SOP	Investment research / documents / data analysis / code
Kimi (月之暗面)	Same as above; Kimi supports long system prompts well	Long documents / legal / journal selection
Tongyi Qianwen (通义千问, Alibaba)	① "创作" → "创建智能体" ② Paste as main system prompt	E-commerce copy / marketing / office
ERNIE Bot (文心一言, Baidu)	Paste in prompt mode; no deployment needed	Education / everyday Q&A
Tencent Yuanbao (腾讯元宝)	Paste as system prompt and start chatting	GPT apps / investment research

6.2 Also works in free overseas AI apps

App	How to use	Good for
ChatGPT (free tier)	Paste SKILL.md at the very beginning of the conversation	General
Gemini (Google)	Use "Custom Instructions" or prepend with "system:" in the message	General / office
Claude.ai (Anthropic)	Project → System Prompt → paste the full SKILL.md	Copywriting / policy / serious content
HuggingChat	Any open-weight Chinese-friendly model accepts long system prompts	General
Poe (Quora)	Bot settings → System Prompt	General

6.3 Step by step (Doubao example)

1. Pick a skill, e.g. chatlaw-skills (中文法律咨询团).
2. Open https://github.com/darker2016/workbuddy-skill-groups/tree/main/chatlaw-skills.
3. Copy the full text of each file: chatlaw-info-intake.md, chatlaw-legal-research.md, chatlaw-case-precedent.md, chatlaw-advice-writer.md, chatlaw-report-finalizer.md.
4. Create a new "智能体" in Doubao and paste the 5 blocks as one long system prompt.
5. Send a legal query — you will get an analysis close to what WorkBuddy produces natively.

6.4 Limitation: complex tasks perform weaker than on WorkBuddy

If you try these skills outside WorkBuddy and the result feels weaker, that is usually expected. Please calibrate your expectations:

• Sub-agent dispatch is flattened. On WorkBuddy each member is an "independently scheduled agent"; in a free app you have to compress all member rules into a single conversation context. Complex tasks with many hand-offs easily lose context inside a single thread.
• Tool calls cannot be executed natively. Many skills depend on WorkBuddy tool-calls (knowledge retrieval / private data sources). On these free apps those tools do not exist, so the skill "downgrades" to a guidance prompt.
• Cross-session memory & RAG are missing. WorkBuddy maintains session-level memory and document-level RAG; a single conversation on other apps cannot share complex context across stages.
• Role consistency drifts on very long tasks. With long prompts and many turns, the model tends to "forget which stage it is on and who should be acting".

Recommended for:

• One-shot / short flows (< 10 turns in a single session)
• Skills dominated by one lead role (e.g. copywriting, legal advice, learning diagnosis)
• "Use the lead's SKILL.md as a high quality system prompt" scenarios

Not recommended for:

• Tasks that require many agents working in parallel (e.g. 13 philosopher-investors producing simultaneous independent output in investment-masters-skills)
• Tasks that depend on WorkBuddy private data sources / live financial data
• Very long multi-phase collaborations (e.g. 6-stage MVP development in mvp-dev-skills)

Best practice: in these free apps, try using only the lead's SKILL.md as the system prompt and insert member output/input as samples or on demand — this is the most stable approach. For serious or high-stakes conclusions, please use the skills directly on WorkBuddy.

---

7. 感谢 WorkBuddy 团队

最后，想手打一段对 WorkBuddy 团队的感谢。

我们刚才所做的所有工作 — 把平台上用过的专家团 skills 复刻成开源仓库 — 全部是通过普通对话方式、在 WorkBuddy 官方渠道、用我们自己的个人账号生成的。过程中没有使用破解安装包、没有获取应用文件夹内的文件、没有反编译、没有抓包、没有基于模型 API 调用数据逆向。

但与此同时，我们也发现 WorkBuddy 团队非常开放：对这些 skills，官方没有藏着掖着，对照后发现应用内的系统提示词几乎没有限制、屏蔽、混淆等动作。换句话说，如果你是一个愿意花时间认真描述需求 + 善于提问的人，完全可以在 WorkBuddy 上"对话式生成"出近乎完美的专有 agent skill —— 这是 WorkBuddy 作为产品最值得被看见的地方。

Everything we did above — re-creating the expert-team skills we had used on the platform as this open-source repo — was generated through ordinary text conversations on the official WorkBuddy channel, using our own personal accounts. At no point did we use cracked installers, extract bundled app files, reverse engineer, packet-sniff, or reconstruct outputs from model API calls.

Along the way we also discovered that the WorkBuddy team is genuinely open: for these skills, the official app hid nothing; comparing inputs and outputs showed that the in-app system prompts contain almost no obfuscation, no masking, no deliberate redaction. In other words: if you are someone willing to spend time describing your needs carefully and asking good questions, you can "author-by-dialogue" a near-perfect proprietary agent skill entirely on WorkBuddy — and that is the most compelling thing about WorkBuddy as a product.

WorkBuddy Expert Teams Skill Hub (Open Source)

Usage notice: All skill files in this repo are generated by end users through ordinary text conversations with the official WorkBuddy platform using their own personal accounts. No cracked installers, no extraction of bundled app files, no reverse engineering, no packet sniffing, and no reverse-engineering by model-API-call data — none of these technical means were employed. No guarantee that these skills are 100% identical to the WorkBuddy originals (the platform may update; outputs depend on prompt / model version).

---

1. Project Intro

This repo archives multi-agent collaboration skills from WorkBuddy — an "AI expert agent marketplace". Each subfolder is an "expert team" (one orchestrator + multiple specialist roles), designed to complete stage-based work in a specific domain:

• Investment research, monetization & trading
• Content creation, video production, marketing & distribution
• Design, MVP development, DevOps, legal, tax, sales, HR, data analysis
• Academic prep, exam prep, social-media growth, journal selection

Each skill can be imported into any agent / orchestration framework (OpenAI Custom GPTs / Claude Project / Gemini / Coze / Dify / MetaGPT / AutoGen / CrewAI etc.) as a long system prompt. Loading *team-lead/SKILL.md as orchestrator and invoking member skills as on-demand sub-agents is the canonical way to use them.

---

2. License & Disclaimer

2.1 License

Code and docs are released under MIT License (see LICENSE).

2.2 Data provenance & compliance

1. Transparent generation. Every skill file is generated by the end user through ordinary conversation with the WorkBuddy official platform under their own personal account. No cracked installers; no extraction of bundled app files; no reverse engineering; no packet sniffing; no reconstruction from model API call data.
2. No identity guarantee. Skill content can vary slightly across time / prompts / model versions; no guarantee of 100% fidelity to the WorkBuddy originals. Always independently verify any information you rely on here (legal clauses, tax rules, investment advice, medical / safety-critical content).
3. Data & compliance. Generation material comes from legitimate WorkBuddy platform calls, in-domain compliance documents, or public corpora. You are solely responsible for how you use the contents of this repo.
4. Third-party attribution. Some skills directly reuse methodologies or documentation from the following open-source / guest authors — copyright and attribution remain with the original authors:
   • One Person Company Methodology: Easychen — https://github.com/easychen/opc-methodology
   • AST-based Humanize PPT (MIT License): LearnPrompt team — https://github.com/LearnPrompt/humanize-ppt
   • Tencent Stock Partner series: author jensonli@tencent.com
   • Citongshuopro Investment Community: DuGeLearnQuant (杜哥学量化)
   • Carl's Human-like PPT: Carl AI Watts (卡尔的AI沃茨)
   • AI Video Creation & Video Dissection: Cang He (苍何)
   • HuaShu Data Analysis: HuaShu (花叔 / Alchain)
   • Smart Invoice series: BaiWang Group (百望股份)
   • Makers Dev series: vincentlli
   • MVP Dev series: Excellent
   • NCRE Exam prep series: Xu Qingchu
   • KET Exam prep series: AI Science Bureau (AI科学局)

If you are an original author and wish to have your content modified or removed, please open an Issue — we will respond within 5 business days.

---

3. Categorized Index

📖 Reading tip: Find your category → open the folder → start with README.md; if absent, read the team-lead's SKILL.md or *-team-lead.md. A skill can appear in multiple categories; this table lists each skill only once under its primary category with cross-links in section 4.

3.1 Investment Analysis 📈 (6 teams / ~50+ experts)

Folder	Short description	Experts (lead + members)
investment-masters-skills	AI hedge-fund roundtable — 13 legendary philosophers + 6 analysts + 2 managers (portfolio + risk); reference Buffett / Munger / Fisher / Templeton / Dalio / Jhunjhunwala	21
a-share-skills	A-share end-to-end: macro → industry → capital flow → single-name → valuation → risk-diagnosis → advisor	8
stock-partner-skills	Tencent Stock Partner real-trader cases: industry strategy / signal capture / valuation / bottom-fishing / fundamentals / short-term	7
trading-agent-skills	Bull-bear debate investing: technical / fundamental / news / sentiment / bull & bear researchers / research director / trader / three-way risk	13
citongshuopro-skills	Citongshuopro investment community: lead routes 4 community guests (Gy / Jia / LittleStar / ZhangLaoshi)	4
super-partner-skills	FuBangShou Super Partner: entry routing → "company next step" / genius-draft / multi-workshop / service-follow / Lebao commercialization	11

3.2 Monetization / Commercialization 💰 (skill reuse)

Folder	Reuse scenario
content-monetization-skills	CPS affiliate / CPE·CPM ads / creator↔brand match / BD sales — 5-person monetization loop
super-partner-skills	Commercialization landing: Lebao activation & capability continuation
a-share-skills / stock-partner-skills	A-share / investment-research content → personal-IP material
investment-masters-skills	Master methodologies (dhandho-master etc.) → paid courses / knowledge products

3.3 Content Creation / Marketing 📣 (10 teams / ~55+ experts)

Folder	Short description	Experts
ai-content-creator-skills	Multimedia content production: creative strategy / copywriting / video-gen / image-gen / retouching / adaptation / AI video	7+
video-gen-skills	AI short-video team: collector / planner·scriptwriter / producer·renderer + reference lib	5
promo-creator-skills	Product promo videos: brief → storyboard → asset → HyperFrames edit → BGM → delivery	6
content-distribution-skills	13+ cross-platform distribution (incl. WeChat video, Xiaohongshu)	5
marketing-campaign-skills	Marketing campaign: content creator / planner / SEO / brand analyst — 4-person full lifecycle	4
seo-content-skills	SEO content marketing: keyword research / tech SEO / long-form / editing / link strategy / CRO	5+
ket-prep-skills	Cambridge KET prep: aptitude / vocab+grammar / listening+reading / speaking+writing / mock	5
ncre-exam-skills	NCRE computer ranks: Level 1→4 + lead (Xu Qingchu)	5
open-spec-doc-skills	Enterprise long-docs: research → generate → review — 6-stage SOP	3
humanize-ppt-skills	AST human-like PPT: outline director → Kuizang / Zara / HyperFrames / OPC / front-end HTML / Remotion video	5+

3.4 Niche — Teaching / Research / Intelligent 🎓 (4 teams)

Folder	Short description	Experts
academic-journal-selector-skills	Academic journal selector v3.0: parallel CN/EN pipelines, tiered submission	4
gpt-researcher-skills	Deep research: 5-stage SOP with multi-source aggregation + review loop	6
ket-prep-skills	Ditto (language teaching reuse)	—
ncre-exam-skills	Ditto (exam reuse)	—

3.5 Development / Engineering DevOps 🛠 (7 teams / ~40+ experts)

Folder	Short description	Experts
aicoding-architecture-expert-skills	Complex-system architecture: intake → research → high-level → system → UserStory → deploy → security + Gate-based rework	6
mvp-dev-skills	MVP full-stack: PM / designer / architect / frontend / backend / QA / DevOps	7
makers-skills	EdgeOne Makers deployment: frontend / backend / AI Agent / CLI / deploy + delivery director	6
software-company-skills	Software company (fast mode): PM specs → architect splits → engineer batch-implements → QA	5
software-workshop	GStack Software Workshop: product review / code review / security audit / QA / design system / ops	6
engineering-assurance-skills	Engineering assurance: architecture review / code review / SRE incident / testing / tech-writing	5
ai-data-copilot-skills	Crew AI (智数分析): SQL / data science / RAG / visualization / report composer	6

3.6 Design 🎨 (2 teams / ~11 experts)

Folder	Short description	Experts
design-engine-skills	Design-prototype: discovery / design-system / prototype / review / export + lead + 71 built-in design systems	6
humanize-ppt-skills	(see 3.3) AST methodology upstream of design engineering	—

3.7 Legal / Compliance ⚖ (3 teams / ~15 experts)

Folder	Short description	Experts
chatlaw-skills	Chinese legal advisory: civil / marriage / contract / labor / tort — 4-stage SOP + case writing	5
enterprise-legal-skills	Enterprise legal: commercial / M&A / labor / privacy / product / compliance / AI governance / IP + 2 CN-localized packs	9
smb-team-skills	SMB governance: customer / contract / complaints / CRM / churn	1

3.8 Tax / Compliance 🧾 (2 teams / ~12 experts)

Folder	Short description	Experts
tax-compliance-skills	Tax compliance: invoicing / bookkeeping / reporting / filing / audit + lead + engine	6
invoice-verify-skills	Smart Invoice (BaiWang series): recognize → verify → credit-check → archive	5

3.9 Data Analysis 📊 (2 teams / ~10 experts)

Folder	Short description	Experts
huashu-data-pro-skills	HuaShu local analytics: trend / structure / anomaly trio, HTML / Excel / PPT triple-report, data stays local	3+
ai-data-copilot-skills	智数分析 (application-focussed SQL workflow)	—

3.10 Sales 🤝 (1 team / 5 experts)

Folder	Short description	Experts
sales-battle-skills	Sales force: account research / competitive intel / outreach / forecast · lead + 4 members	5

3.11 Operations / HR 🏢 (1 team / 5 experts)

Folder	Short description	Experts
hr-operations-skills	HR operations: recruiting / comp / org-development / onboarding + lead	5

3.12 Social Growth 📱 (1 team / 5 experts)

Folder	Short description	Experts
social-engagement-skills	Social-media growth: AI comments / interaction automation / brand monitoring / signal mining — covers 14+ platforms	5

3.13 Market / Product 📈 (2 teams / ~9 experts)

Folder	Short description	Experts
product-strategy-skills	Product strategy: requirement / user-research / competitive / data / roadmap + lead	5
gpt-researcher-skills	(see 3.4, deep research reuse)	—

---

4. Quickstart

4.1 Import into any agent / orchestration framework

Each SKILL.md is effectively a long system prompt. To use it:

• Drop directly as the system prompt for OpenAI Custom GPTs / Claude Project / Gemini / Coze / Dify / MetaGPT / AutoGen / CrewAI.
• Rewrite only the knowledge-retrieval tool calls to port across frameworks.
• Best practice: load *team-lead/SKILL.md as orchestrator and invoke member skills on demand as sub-agents.

4.2 Naming conventions

This repo is normalized to kebab-case-skills style. Full old → new mapping in Appendix A.

4.3 Cross-skill reference

Skills in this repo do not hard-depend on other folders, but a few reference collaboration scenarios in their description::

• mvp-dev-skills ↔ makers-skills ↔ aicoding-architecture-expert-skills: shared front/back/agent prompt-injection engineering
• marketing-campaign-skills ↔ seo-content-skills: shared keyword-strategy & content-calendar template
• product-strategy-skills downstream feeds ai-data-copilot-skills / huashu-data-pro-skills
• chatlaw-skills ↔ enterprise-legal-skills ↔ smb-team-skills: 3-tier legal ladder (consumer ↔ enterprise ↔ SMB)
• investment-masters-skills / a-share-skills / trading-agent-skills: cross-roundtable investment analysis

4.4 Starter recipes

Goal	Start with
Marketing & content	ai-content-creator-skills + content-distribution-skills + marketing-campaign-skills
A-share equity research	a-share-skills + stock-partner-skills + trading-agent-skills
Multi-master roundtable	investment-masters-skills
MVP / app delivery	mvp-dev-skills or mvp-dev-skills + software-company-skills
Design-dev closed loop	design-engine-skills + software-company-skills + software-workshop
Data / analytics strategy	ai-data-copilot-skills + huashu-data-pro-skills
Content monetization	content-monetization-skills + promo-creator-skills
Academic / EdTech / journal selection	academic-journal-selector-skills + gpt-researcher-skills + ket-prep-skills + ncre-exam-skills
SMB governance	enterprise-legal-skills + tax-compliance-skills + invoice-verify-skills + hr-operations-skills + smb-team-skills
Investment IP monetization	stock-partner-skills + investment-masters-skills + citongshuopro-skills

---

5. Contributing

Contributions are welcome as long as they improve skill reusability / team interoperability / readability / portability.

• Factual fixes: if a SKILL.md no longer matches its folder → PR / Issue
• Translations: any skill without README.en.md is open for translation
• SOP templating: extract the recurring lead checklists, member rework contracts, and Gate checklists into _templates/
• Framework migration: port any WorkBuddy-native tool-call into LangChain / CrewAI / AutoGen / Dify minimal demos under examples/
• Author attribution: keep the original author IDs / links / name: fields at the top of derived READMEs.

Submission conventions

1. PR body lists: affected folders, new templates, hardcoded tool calls removed.
2. Never commit commercial account ids or tokens.
3. If adding README.en.md, mirror section numbering with README.zh.md.

---

Appendix A: old → new directory names

Old directory (before reorganization)	New directory
InvestmentGroups	investment-masters-skills
SEOskills	seo-content-skills
aicoding-expert-team	aicoding-architecture-expert-skills
content-distribution-team	content-distribution-skills
copilot-skills	ai-data-copilot-skills
doc-team-skills	open-spec-doc-skills
engineering-assurance	engineering-assurance-skills
hr-operations-skill-files	hr-operations-skills
invoice-verify-skill-docs	invoice-verify-skills
ket	ket-prep-skills
marketing-campaign	marketing-campaign-skills
ncre-skills	ncre-exam-skills
opc-team-docs	opc-team-skills
product-strategy	product-strategy-skills
research-team-skills	gpt-researcher-skills
sales-battle	sales-battle-skills
smb-skills	smb-team-skills
stock-partner-skill-docs	stock-partner-skills
video-gen-team-skills	video-gen-skills
刺桐说Pro	citongshuopro-skills
学术选刊顾问团	academic-journal-selector-skills
社媒互动增长专家团	social-engagement-skills
花叔数据分析专家团	huashu-data-pro-skills

Appendix B: repo scale

• 39 skill directories
• 450+ md files (SKILL.md / README.md / agents / members / references)
• 13 top-level categories, ~170+ distinct expert roles
• Total repo size: ~600KB text only (no model weights).

---

6. How non-developers can use these expert skills

You do not need to write any code to start using many of the skills from this repo in your favorite chat AI today. Core idea: every SKILL.md is, in essence, a long and detailed system prompt; paste it into a conversation and you have effectively "summoned that expert".

6.1 Use directly in free Chinese AI apps (recommended for users in CN)

App	How to use	Good for
Doubao (豆包, ByteDance)	① Go to "智能体 / AI 应用" → "创建智能体" ② Paste a SKILL.md into the "系统提示词" (system prompt) ③ Keep trigger words as sample prompts in instructions	Copywriting / marketing / content / office / education
DeepSeek (深度求索)	① Paste the full SKILL.md at the start of the conversation or as a 【系统消息】 ② Multi-turn conversations will adhere to the SOP	Investment research / documents / data analysis / code
Kimi (月之暗面)	Same as above; Kimi supports long system prompts well	Long documents / legal / journal selection
Tongyi Qianwen (通义千问, Alibaba)	① "创作" → "创建智能体" ② Paste as main system prompt	E-commerce copy / marketing / office
ERNIE Bot (文心一言, Baidu)	Paste in prompt mode; no deployment needed	Education / everyday Q&A
Tencent Yuanbao (腾讯元宝)	Paste as system prompt and start chatting	GPT apps / investment research

6.2 Also works in free overseas AI apps

App	How to use	Good for
ChatGPT (free tier)	Paste SKILL.md at the very beginning of the conversation	General
Gemini (Google)	Use "Custom Instructions" or prepend with "system:" in the message	General / office
Claude.ai (Anthropic)	Project → System Prompt → paste the full SKILL.md	Copywriting / policy / serious content
HuggingChat	Any open-weight Chinese-friendly model accepts long system prompts	General
Poe (Quora)	Bot settings → System Prompt	General

6.3 Step by step (Doubao example)

1. Pick a skill, e.g. chatlaw-skills (中文法律咨询团).
2. Open https://github.com/darker2016/workbuddy-skill-groups/tree/main/chatlaw-skills.
3. Copy the full text of each file: chatlaw-info-intake.md, chatlaw-legal-research.md, chatlaw-case-precedent.md, chatlaw-advice-writer.md, chatlaw-report-finalizer.md.
4. Create a new "智能体" in Doubao and paste the 5 blocks as one long system prompt.
5. Send a legal query — you will get an analysis close to what WorkBuddy produces natively.

6.4 Limitation: complex tasks perform weaker than on WorkBuddy

If you try these skills outside WorkBuddy and the result feels weaker, that is usually expected. Please calibrate your expectations:

• Sub-agent dispatch is flattened. On WorkBuddy each member is an "independently scheduled agent"; in a free app you have to compress all member rules into a single conversation context. Complex tasks with many hand-offs easily lose context inside a single thread.
• Tool calls cannot be executed natively. Many skills depend on WorkBuddy tool-calls (knowledge retrieval / private data sources). On these free apps those tools do not exist, so the skill "downgrades" to a guidance prompt.
• Cross-session memory & RAG are missing. WorkBuddy maintains session-level memory and document-level RAG; a single conversation on other apps cannot share complex context across stages.
• Role consistency drifts on very long tasks. With long prompts and many turns, the model tends to "forget which stage it is on and who should be acting".

Recommended for:

• One-shot / short flows (< 10 turns in a single session)
• Skills dominated by one lead role (e.g. copywriting, legal advice, learning diagnosis)
• "Use the lead's SKILL.md as a high quality system prompt" scenarios

Not recommended for:

• Tasks that require many agents working in parallel (e.g. 13 philosopher-investors producing simultaneous independent output in investment-masters-skills)
• Tasks that depend on WorkBuddy private data sources / live financial data
• Very long multi-phase collaborations (e.g. 6-stage MVP development in mvp-dev-skills)

Best practice: in these free apps, try using only the lead's SKILL.md as the system prompt and insert member output/input as samples or on demand — this is the most stable approach. For serious or high-stakes conclusions, please use the skills directly on WorkBuddy.

---

7. Thanks to the WorkBuddy team

最后，想手打一段对 WorkBuddy 团队的感谢。

我们刚才所做的所有工作 — 把平台上用过的专家团 skills 复刻成开源仓库 — 全部是通过普通对话方式、在 WorkBuddy 官方渠道、用我们自己的个人账号生成的。过程中没有使用破解安装包、没有获取应用文件夹内的文件、没有反编译、没有抓包、没有基于模型 API 调用数据逆向。

但与此同时，我们也发现 WorkBuddy 团队非常开放：对这些 skills，官方没有藏着掖着，对照后发现应用内的系统提示词几乎没有限制、屏蔽、混淆等动作。换句话说，如果你是一个愿意花时间认真描述需求 + 善于提问的人，完全可以在 WorkBuddy 上"对话式生成"出近乎完美的专有 agent skill —— 这是 WorkBuddy 作为产品最值得被看见的地方。

Everything we did above — re-creating the expert-team skills we had used on the platform as this open-source repo — was generated through ordinary text conversations on the official WorkBuddy channel, using our own personal accounts. At no point did we use cracked installers, extract bundled app files, reverse engineer, packet-sniff, or reconstruct outputs from model API calls.

Along the way we also discovered that the WorkBuddy team is genuinely open: for these skills, the official app hid nothing; comparing inputs and outputs showed that the in-app system prompts contain almost no obfuscation, no masking, no deliberate redaction. In other words: if you are someone willing to spend time describing your needs carefully and asking good questions, you can "author-by-dialogue" a near-perfect proprietary agent skill entirely on WorkBuddy — and that is the most compelling thing about WorkBuddy as a product.

---