给 AI Agent 装一个 CMO 大脑：marketingskills 开源营销技能包
Give Your AI Agent a CMO Brain: marketingskills Open-Source Marketing Skill Pack
结论先行（BLUF）：你的 AI Agent 会写代码，但不会做增长。marketingskills 把资深营销专家 Corey Haines 的方法论打包成 Agent 可执行的技能规范，32+ 个专业营销技能 + 51 个 CLI 工具直连真实 API，装完就是 CMO 级别的视野。GitHub：coreyhaines31/marketingskills
它解决了什么问题
AI Agent 在编程上已经接近专家水平，但面对”如何让产品获客""如何提升注册转化""如何做 SEO”这些问题时，往往只能给出教科书式的泛泛之谈，缺乏可落地的专业深度。
marketingskills 的逻辑是：把专家知识结构化为 Agent 可理解的指令集。不是让 AI 去搜索营销知识，而是直接在仓库里给 Agent 装好”专业大脑”。
核心功能模块
1. AI 时代 SEO 策略（含 AEO/GEO）
不只是传统的技术 SEO 审计，而是重点引入了 AI SEO（AEO/GEO） 框架：
- 如何优化内容以在 ChatGPT、Perplexity、Gemini 等生成式搜索引擎中被引用
- BLUF 结构、FAQ 模块、结构化数据标记等 GEO 战术
- 传统排名信号 + AI 引用信号双轨并行策略
2. 全链路 CRO 转化率优化
从落地页架构 → 注册流设计 → 流失预防的完整框架：
- 不只告诉 AI “写什么”，通过心理学模型解释”为什么这样布局”
- 涵盖首屏信息优先级、社会证明摆放位置、行动号召措辞等细节
- 注册流每个步骤的摩擦点识别与消除策略
3. 数据驱动文案生成
内置多种高转化文案框架，重点解决 AI 腔调 问题：
- 具体的文案结构（AIDA、PAS、Before-After-Bridge 等）
- 广告素材创作：Facebook/Google 广告、冷邮件序列、落地页标题
- 产品描述、定价页文案、用户评价展示方式
4. 增长工程与 RevOps
覆盖整个收入运营链路：
- 潜在客户评分模型与销售赋能工具
- 51 个零依赖 CLI 工具，Agent 可直接调用：
- Ahrefs API → 关键词分析、竞品反链
- Google Ads API → 广告效果数据
- Mailchimp API → 邮件序列自动化
- 更多营销平台集成
技术架构
遵循 .agents/skills/ 跨 Agent 标准，和本博客使用的技能规范格式完全兼容：
# 快速安装
npx skills install coreyhaines31/marketingskills“知识 + 工具”双层结构：
SKILL.md— 营销方法论，Agent 推理时读取- CLI 工具 — 实时 API 数据，Agent 执行时调用
.agents/product-marketing-context.md— 产品背景注入，让建议有针对性
Agent 不是在讲通用营销理论，而是基于你的产品具体情况给出可执行建议，并能直接拉取真实数据验证。
对开发者的实际意义
| 场景 | 没有 marketingskills | 有 marketingskills |
|---|---|---|
| 产品上线前 | AI 给出模糊建议 | 完整落地页 CRO 审计 + 文案框架 |
| SEO 优化 | 通用 checklist | AI SEO/GEO 双轨策略 + 关键词数据 |
| 写营销文案 | 充满”AI 腔”的套话 | 有心理学依据的高转化结构 |
| 获客增长 | 无从下手 | RevOps 框架 + 实时 API 数据支撑 |
快速上手
# 安装技能包
npx skills install coreyhaines31/marketingskills
# 在 Claude Code 中直接调用
# /marketingskills:cro-audit → CRO 转化审计
# /marketingskills:seo-strategy → SEO/GEO 优化策略
# /marketingskills:copy-review → 文案质量检查配合 .agents/product-marketing-context.md 填入产品信息后，Agent 的建议会从通用框架变成针对你产品的具体执行方案。
常见问题
Q：这和普通的营销 Prompt 模板有什么区别？
A：核心区别是”可执行性”。普通 Prompt 只能指导 AI 生成文字，marketingskills 的 CLI 工具层让 Agent 可以主动拉取真实数据（关键词搜索量、竞品数据、邮件打开率），做到数据驱动而非拍脑袋。
Q：不用 Claude Code，其他 Agent 也能用吗？
A：支持任何遵循 .agents/skills/ 规范的 Agent 框架。Cursor、Windsurf 等理论上也可集成，前提是支持自定义 skill 规范。
Q：32+ 技能一次性装进上下文不会太重吗？
A：技能按需加载，不是一次性全部注入。Agent 根据任务类型调用对应的 SKILL.md，其他技能不占用上下文窗口。
Q：AEO/GEO 优化真的有效果吗？
A：在 AI 搜索份额快速增长的当下，GEO 优化已经出现可量化的引用率提升案例。marketingskills 的 GEO 模块和本博客自建的 SEO/GEO 技能思路一致：BLUF 结构、FAQ 模块、品牌实体标注是核心战术。
Q：51 个 CLI 工具需要自己配置 API Key 吗？
A：是的，每个工具需要对应平台的 API 授权。工具本身零依赖（无 npm 包），只需提供 Key 即可运行。
© 2026 Author: Mycelium Protocol. 本文采用 CC BY 4.0 授权——欢迎转载和引用，须注明作者姓名及原文链接，不得去除署名后以原创发布。
BLUF: Your AI agent knows how to code but doesn’t know how to grow. marketingskills packages senior marketer Corey Haines’ methodology into agent-executable skill specs: 32+ professional marketing skills + 51 CLI tools with live API access. Install it and your agent has CMO-level expertise. GitHub: coreyhaines31/marketingskills
The Problem It Solves
AI agents are approaching expert-level coding ability, but when asked “how do we acquire users,” “how do we improve sign-up conversion,” or “what should our SEO strategy be” — they produce textbook generalities with no actionable depth.
marketingskills takes a different approach: structure expert knowledge into agent-executable instruction sets. Don’t make the agent search for marketing knowledge — give it a professional brain directly in the repo.
Core Modules
1. AI-Era SEO Strategy (AEO/GEO)
Beyond traditional technical SEO audits, the project introduces AI SEO (AEO/GEO) frameworks:
- How to optimize content for citation by ChatGPT, Perplexity, and Gemini
- GEO tactics: BLUF structure, FAQ modules, structured data markup
- Dual-track strategy: traditional ranking signals + AI citation signals in parallel
2. Full-Funnel CRO
A complete framework from landing page architecture → signup flow → churn prevention:
- Explains not just “what to write” but “why this layout” via psychological models
- Covers above-the-fold information hierarchy, social proof placement, CTA wording
- Friction point identification and removal at each step of the signup flow
3. Data-Driven Copywriting
Multiple high-converting copy frameworks with a specific focus on eliminating AI tone:
- Structured frameworks: AIDA, PAS, Before-After-Bridge
- Ad creative for Facebook/Google, cold email sequences, landing page headlines
- Product descriptions, pricing page copy, testimonial display strategies
4. Growth Engineering & RevOps
Full revenue operations coverage:
- Lead scoring models and sales enablement tools
- 51 zero-dependency CLI tools callable by agents:
- Ahrefs API → keyword analysis, competitor backlinks
- Google Ads API → campaign performance data
- Mailchimp API → email sequence automation
- Additional marketing platform integrations
Technical Architecture
Follows the .agents/skills/ cross-agent standard — fully compatible with the skill spec format used in this blog:
npx skills install coreyhaines31/marketingskillsTwo-layer “knowledge + tools” structure:
SKILL.md— marketing methodology for agent reasoning- CLI tools — live API data for agent execution
.agents/product-marketing-context.md— product context injection for targeted advice
The agent isn’t delivering generic marketing theory — it’s working from your specific product context, with real data to back the recommendations.
Practical Impact for Developers
| Scenario | Without marketingskills | With marketingskills |
|---|---|---|
| Pre-launch | Vague AI suggestions | Full CRO audit + copy frameworks |
| SEO | Generic checklist | Dual-track AI SEO/GEO + keyword data |
| Copywriting | AI-sounding boilerplate | Psychologically-grounded high-converting structure |
| Growth | No direction | RevOps framework + live API data |
Quick Start
npx skills install coreyhaines31/marketingskills
# Then in Claude Code:
# /marketingskills:cro-audit
# /marketingskills:seo-strategy
# /marketingskills:copy-reviewFill in .agents/product-marketing-context.md with your product details and the agent’s advice shifts from generic frameworks to specific, executable recommendations for your product.
FAQ
Q: How is this different from regular marketing prompt templates?
A: The core difference is executability. Prompt templates only generate text. The CLI tool layer in marketingskills lets the agent pull live data — keyword volumes, competitor analysis, email open rates — making recommendations data-driven rather than opinion-based.
Q: Does it work with agents other than Claude Code?
A: Any agent framework that supports the .agents/skills/ spec. Cursor and Windsurf are theoretically compatible, as long as they support custom skill specs.
Q: Won’t loading 32+ skills overwhelm the context window?
A: Skills load on demand — not all injected at once. The agent calls the relevant SKILL.md for each task; other skills don’t consume context.
Q: Does AEO/GEO optimization actually produce measurable results?
A: With AI search share growing rapidly, early GEO optimization cases show quantifiable citation rate improvements. The GEO module in marketingskills aligns with the tactics in this blog’s own SEO/GEO skill: BLUF structure, FAQ modules, and brand entity tagging are the core levers.
Q: Do the 51 CLI tools require API key setup?
A: Yes — each tool needs the corresponding platform’s API authorization. The tools themselves are zero-dependency (no npm packages), requiring only your key to run.
© 2026 Author: Mycelium Protocol. Licensed under CC BY 4.0 — free to share and adapt with attribution. You must credit the author and link to the original; removing attribution and republishing as original is not permitted.
💬 评论与讨论
使用 GitHub 账号登录后发表评论