Skip to content

fullstackcrew-alpha/skill-cn-content-matrix

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

📱 CN Content Matrix

One topic, four platforms, zero reformatting. True style transfer for Chinese social media.

GitHub stars OpenClaw Skill License: MIT Version Platforms

If you find this useful, please give it a star — it helps others discover it.

Overview

CN Content Matrix is not a format converter — it's a style transfer engine for Chinese social media. Given a single topic, it generates platform-native content that reads like it was written by a real KOL on each platform, not by an AI doing find-and-replace on templates.

The core insight: a Xiaohongshu post about "home coffee brewing" should feel like your best friend gushing about a discovery, while the same topic on WeChat Official Account should read like a thoughtful industry analysis. Same information, completely different voice, structure, and emotional appeal.

Supported Platforms

Platform Code Word Count Tone Key Elements
小红书 (Xiaohongshu) xhs 300–800 Casual, bestie-sharing, seed-planting Emoji-dense, hashtags, first-person sharing, short paragraphs (≤3 sentences)
微信公众号 (WeChat MP) wechat 1,500–8,000 Professional, insightful, authoritative Long-form analysis, data-driven, structured sections, SEO-optimized
抖音 (Douyin/TikTok CN) douyin 200–5,000 Fast-paced, punchy, high-energy Script format, 15-char max per line, 3-second hook, scene cuts annotated
B站 (Bilibili) bilibili 3,000–15,000 Knowledgeable, meme-savvy, friendly expert Rich text (code blocks, tables), table of contents, deep-dive structure

Features

  • True Style Transfer — persona-switching, not template-filling. Each platform gets content written in its native voice
  • Content Core Architecture — all platform variants derive from a shared "Content Core" document, ensuring message consistency with style diversity
  • Real-time Research — WebSearch-powered topic research with current-year trending angles and keywords
  • Sensitive Word Detection — three-tier screening (🔴 hard-ban / 🟡 soft-limit / 🟢 caution) against Chinese advertising law and platform rules
  • Content Compliance Review — 6-dimension scoring system (word count, format, tone, sensitive words, SEO, publishability) with 60-point scale
  • Shell-based Quality Checkcontent-check.sh script for quick automated compliance validation
  • Anti-AI Detection — built-in rules to eliminate "AI smell" phrases like "随着科技的发展", "值得注意的是", etc.
  • Cross-platform Differentiation Check — automated detection of content overlap (>50 chars repeated = fail)
  • Platform-specific CTAs — Xiaohongshu → save+comment, WeChat → follow+share, Douyin → comment+follow, Bilibili → triple-click

Quick Start

# Install: add to your Claude Code skill directory
# Then invoke via slash command:

# Generate content for a single platform (defaults to Xiaohongshu)
/cn-content-matrix content-gen 家用咖啡机选购 xhs

# Generate for all 4 platforms at once
/cn-content-matrix content-matrix 家用咖啡机选购

# Review existing content for compliance
/cn-content-matrix content-review xhs ~/content-output/2026-03-21/coffee/xhs.md

Output is saved to ~/content-output/{date}/{topic-slug}/.

Commands

/cn-content-matrix content-gen [topic] [platform]

Generate content for a single platform.

# Examples
/cn-content-matrix content-gen AI编程工具推荐 wechat
/cn-content-matrix content-gen 春季护肤 xhs
/cn-content-matrix content-gen 副业赚钱 douyin
/cn-content-matrix content-gen Rust语言入门 bilibili

Pipeline: Topic Research → Load Platform Spec → Generate with Style Transfer → Self-check → Output

Output file: ~/content-output/{date}/{topic}/{platform}.md

# Output frontmatter
---
platform: 小红书
topic: 春季护肤
word_count: 526
generated_at: 2026-03-21T14:30:00+08:00
self_check: PASS
---

/cn-content-matrix content-matrix [topic]

Generate a full content matrix — all 4 platforms from a single Content Core.

/cn-content-matrix content-matrix 家用咖啡机选购
/cn-content-matrix content-matrix 2026年AI趋势

Pipeline: Research → Build Content Core → Generate in order (WeChat → Bilibili → Xiaohongshu → Douyin) → Differentiation Check → Overview Report

Output structure:

~/content-output/{date}/{topic}/matrix/
├── content-core.md    # Shared content foundation
├── wechat.md          # WeChat Official Account version
├── bilibili.md        # Bilibili column version
├── xhs.md             # Xiaohongshu version
├── douyin.md          # Douyin script version
└── overview.md        # Cross-platform comparison table

The overview.md includes a comparison table across all platforms (word count, tone, angle, selling points, CTA, special elements) plus publishing time recommendations.

/cn-content-matrix content-review [platform] [file]

Review existing content for platform compliance. Scores across 6 dimensions on a 60-point scale.

/cn-content-matrix content-review xhs ./my-draft.md
/cn-content-matrix content-review wechat ./article.md

Output: A detailed review report with:

  • Overall verdict: PASS ✅ / WARNING ⚠️ / FAIL ❌
  • Per-dimension scores (0–10 each): word count, format structure, tone match, sensitive words, SEO friendliness, publishability
  • Prioritized fix list: 🔴 must-fix → 🟡 should-fix → 🟢 nice-to-have
  • For FAIL results: auto-generated fix-suggestions-{platform}.md with copy-paste replacements

Style Transfer Engine

This is the core differentiator. Here's the same topic — "Home Coffee Brewing" — adapted for each platform:

Xiaohongshu (bestie sharing)

☕ 姐妹们!我终于找到在家做出咖啡店味道的方法了!

之前花了大几千买的咖啡机根本不好用 😭
直到我发现了这个 300 块的手冲组合...

✅ 磨豆机选这款就够了
✅ 滤杯别买贵的,新手用扇形
✅ 水温 92 度是灵魂

你们觉得呢?评论区告诉我你们平时怎么冲 ☕
#家用咖啡 #手冲咖啡 #咖啡入门

WeChat Official Account (industry insight)

当精品咖啡从「第三空间」走进「第一空间」

据美团数据显示,2025年中国家庭咖啡消费同比增长47%。
这不仅是消费习惯的迁移,更是一场静默的品质革命...

本文将从设备选择、豆源甄别、萃取参数三个维度,
为你搭建一套可复用的家庭咖啡体系。

Douyin (3-second hook script)

[画面:精美拉花特写]
「千万别买咖啡机」

[字幕强调:千万别买]
「在你看完这条视频之前」

[画面切换:一堆闲置咖啡机]
「因为我踩过的坑,能帮你省3000块」

[口播加速]
「300块,3样东西,在家做出店里的味道」

Bilibili (knowledgeable friend)

# 从零搭建家用咖啡系统 —— 一个软件工程师的咖啡装备进化史

## 目录
1. 为什么我放弃了胶囊机
2. 手冲 vs 意式 vs 摩卡壶:参数化对比
3. 300元档 vs 3000元档盲测结果
4. 我的最终方案(附购买链接)

> 前方高能预警:大量数据表格和对比图
> 一键三连支持一下,评论区见~

Platform Specifications

Each platform has a comprehensive spec file in references/:

File Covers
references/xiaohongshu.md Title formulas (6 types), body structure, emoji/hashtag rules, engagement patterns, algorithm preferences
references/wechat-mp.md Title strategies, long-form structure, SEO optimization for 搜一搜, subscription vs recommendation traffic
references/douyin.md 7 hook formulas for the critical 3-second window, script format, scene annotation, pacing rules
references/bilibili.md Rich text formatting (code blocks, tables), community culture (弹幕, 梗), title strategies for tech content
references/content-core-template.md Template for building the shared Content Core (key arguments, data points, audience personas, emotional tone, keywords)
references/sensitive-words.md Three-tier sensitive word database with replacement suggestions, updated for 2026

Content Quality Check

The scripts/content-check.sh script provides fast, automated compliance validation from the command line:

./scripts/content-check.sh <file-path> <platform-code>

# Example
./scripts/content-check.sh ~/content-output/2026-03-21/coffee/xhs.md xhs

Checks performed:

Check Details
Word Count Validates against platform-specific ranges (e.g., xhs: 300–800, wechat: 1500–8000)
Sensitive Words Scans for 🔴 hard-ban terms (advertising law violations, fake claims, traffic-diversion words)
Emoji Density Platform-appropriate emoji count (xhs: 5–20 required, wechat: ≤10 recommended)
Paragraph Length Flags overly long paragraphs (xhs: >200 chars, wechat: >800 chars, bilibili: >600 chars)

Exit codes: 0 = PASS, 1 = WARNING, 2 = FAIL

Sample output:

============================================
 内容合规检查报告
 平台: 小红书 (xhs)
 文件: ~/content-output/coffee/xhs.md
============================================

总评: PASS

--- 逐项检查 ---

[字数]       PASS       字数 526,范围 300-800
[敏感词]     PASS       未发现硬违禁词
[Emoji]      PASS       Emoji 12 个(推荐 5-20)
[段落]       PASS       最长段落 156 字符(OK)

Sensitive Word Detection

Three-tier classification system based on Chinese advertising law and platform-specific rules:

Level Meaning Action Examples
🔴 Hard Ban Violates advertising law or triggers instant content removal Must remove or replace 最好, 最佳, 100%, 保证, 根治, 稳赚不赔, 日入过万
🟡 Soft Limit May trigger algorithmic throttling / reduced distribution Recommended to replace Platform-specific marketing terms, borderline claims
🟢 Caution Context-dependent, may be ambiguous Review case-by-case Terms that are fine in some contexts but risky in others

Each banned term comes with suggested replacements (e.g., "最好" → "很不错", "100%" → "大概率").

Platform-specific additions:

  • Xiaohongshu: traffic diversion terms (链接在评论区, 暴富, 月入)
  • WeChat: viral bait (转发领红包, 不转不是中国人)
  • Douyin: income claims (赚钱, 搞钱)

Output Structure

All generated content follows a consistent output format:

---
platform: {platform name}
topic: {topic}
word_count: {count}
generated_at: {ISO 8601 timestamp}
self_check: PASS / WARNING
warnings: [{list of warnings, if any}]
---

{publishable content body}

---
<!-- 以下为生成器备注,发布时删除 -->
{generator notes, review warnings, modification suggestions}

Content is saved to: ~/content-output/{YYYY-MM-DD}/{topic-slug}/

If the same topic is generated multiple times on the same day, a timestamp suffix is added to avoid overwriting.

Contributing

Contributions are welcome! Areas where help is appreciated:

  • Platform spec updates — Chinese social media rules change frequently; help keep references/ current
  • Sensitive word database — additions, removals, and replacement suggestions
  • New platform support — Zhihu, Weibo, Kuaishou, etc.
  • content-check.sh enhancements — additional automated checks

License

MIT


中文说明

项目简介

CN Content Matrix 是一个基于 Claude Code 的中文多平台内容矩阵生成器。它的核心理念是风格迁移而非格式转换——同一个主题,用四种完全不同的"人格"来表达,让每个平台的内容都像是该平台的原生 KOL 写的。

为什么需要这个工具?

做过内容矩阵的人都知道痛点:

  • ❌ 把公众号文章删减一下发小红书 → 读起来像说明书
  • ❌ 把小红书笔记拉长一下发公众号 → 读起来像聊天记录
  • ❌ 一稿多发 → 每个平台都不温不火

CN Content Matrix 的解决方案: 不是改格式,而是换人格。

平台 你的人格身份 你在对谁说话 写作风格
小红书 热爱生活的分享达人 闺蜜/同好 第一人称、种草体、emoji 密集、口语化
微信公众号 有深度的行业观察者 关注专业内容的读者 深度分析、数据论证、引经据典、书面语
抖音 说话有趣的内容创作者 刷到你的路人 口播脚本、每句≤15字、前3秒强钩子
B站 懂行的知识UP主 愿意看深度内容的粉丝 技术科普、用梗、弹幕文化、代码块支持

三大核心功能

1. 单平台内容生成 (content-gen)

给定主题和目标平台,生成一篇可直接复制发布的平台原生内容。

/cn-content-matrix content-gen 春季护肤攻略 xhs

执行流程:

  1. 主题调研 — 通过 WebSearch 获取最新趋势、热搜关键词、竞品切入点
  2. 加载平台规范 — 读取对应平台的完整创作规范和敏感词库
  3. 风格迁移生成 — 以该平台 KOL 的身份和语气撰写内容
  4. 自查自纠 — 字数、格式、语气、敏感词、原创性、可发布性六项检查
  5. 输出文件 — 带 YAML frontmatter 的 Markdown 文件,可直接复制发布

平台代号对照:

代号 平台 默认
xhs 小红书 ✅(未指定平台时默认)
wechat 微信公众号
douyin 抖音
bilibili / bili B站专栏

2. 全平台矩阵生成 (content-matrix)

一个主题,同时生成四个平台的差异化内容。这是本工具的核心功能。

/cn-content-matrix content-matrix 2026年AI编程趋势

执行流程:

  1. 构建内容核心(Content Core) — 提炼核心论点、关键数据、受众画像、情绪基调、关键词集
  2. 主题调研 — 多维度 WebSearch 获取最新信息
  3. 按序生成四平台内容:
    • 微信公众号(最深度,先生成作为素材库)
    • B站专栏(次深度,复用分析但调整语气)
    • 小红书(从深度内容提炼种草点)
    • 抖音脚本(提炼最抓眼球的信息)
  4. 差异化检查 — 确保各平台开头不同、论证结构不同、无大段重复(>50字连续相同=不合格)、CTA适配平台
  5. 生成内容一览表 — 四平台横向对比 + 发布时间建议

输出目录结构:

~/content-output/2026-03-21/ai编程趋势/matrix/
├── content-core.md    # 内容核心(所有平台的共同基础)
├── wechat.md          # 微信公众号版本
├── bilibili.md        # B站专栏版本
├── xhs.md             # 小红书版本
├── douyin.md          # 抖音脚本版本
└── overview.md        # 内容一览表(四平台对比 + 发布建议)

3. 内容合规审查 (content-review)

对已有内容进行平台合规性审查,60分满分制,6个维度各10分。

/cn-content-matrix content-review xhs ./my-xiaohongshu-draft.md

六维评分体系:

维度 检查内容 满分
字数 是否在平台规范范围内 10
格式结构 标题格式、段落结构、平台特有元素(emoji/hashtag/时间轴等) 10
语气匹配 是否符合平台调性,有无"串台"现象 10
敏感词 三级敏感词扫描(🔴🟡🟢) 10
SEO友好度 关键词密度2%-5%、标题含核心词、hashtag相关性 10
可发布性 是否可直接复制发布、有无残留占位符 10

及格线:36/60。 FAIL 结果会自动生成 fix-suggestions-{platform}.md,包含所有🔴级问题的具体修改文本。

敏感词检测详解

三级分类系统,基于《广告法》及各平台最新规则:

🔴 硬违禁(必须删除或替换):

  • 绝对化用语:最好、最佳、最强、最大、最低价、唯一、100%、绝对
  • 虚假宣传:根治、治愈、月瘦X斤、稳赚不赔、日入过万
  • 引流违禁:微信号、vx、wx、加我、私我、淘口令
  • 各平台特有违禁词(如小红书禁"链接在评论区",抖音禁"赚钱/搞钱")

🟡 软限制(可能触发限流):

  • 平台特定的营销敏感词
  • 擦边表述

🟢 注意(视语境处理):

  • 在特定上下文可能产生歧义的词汇

每个违禁词都提供了替代表达(例如:"最好" → "很不错","100%" → "大概率")。

命令行合规检查工具

除了 AI 驱动的 content-review 命令,还提供了一个 shell 脚本用于快速自动化检查:

# 用法
./scripts/content-check.sh <文件路径> <平台代号>

# 示例
./scripts/content-check.sh ~/content-output/2026-03-21/coffee/xhs.md xhs

检查项目:字数范围、硬违禁词扫描、Emoji 密度(小红书要求5-20个)、段落长度。

退出码:0 = 通过,1 = 警告,2 = 不通过。可集成到 CI/CD 或发布前自动检查流程。

反 AI 味机制

以下表达在任何平台都被禁止使用:

"在当今社会"、"随着科技的发展"、"众所周知"、"值得注意的是"、"不可否认"、"总而言之"、"让我们一起来看看"、过于工整的排比句、每段都差不多长的"均匀体"

替代原则: 用该平台真实用户会用的表达。宁可口语化到略显粗糙,也不要光滑到像 AI。

安装方式

cn-content-matrix 目录放入你的 Claude Code skill 目录即可:

# 克隆仓库
git clone <repo-url>

# 将 skill 目录复制或链接到你的 skills 目录
cp -r cn-content-matrix /path/to/your/skills/

# 验证安装
# 在 Claude Code 中输入:
/cn-content-matrix content-gen 测试主题 xhs

各平台规范文件说明

文件 内容覆盖 更新频率
references/xiaohongshu.md 标题公式(6种)、正文结构、emoji/hashtag规范、互动引导、算法偏好 季度更新
references/wechat-mp.md 标题策略、长文结构、搜一搜SEO优化、订阅vs推荐流量 季度更新
references/douyin.md 7种钩子公式(前3秒生死线)、脚本格式、画面标注、节奏控制 季度更新
references/bilibili.md 富文本格式(代码块/表格)、社区文化(弹幕/梗)、技术内容标题策略 季度更新
references/sensitive-words.md 三级敏感词库 + 替代表达建议 持续更新
references/content-core-template.md 内容核心模板(论点/数据/受众/情绪/关键词) 稳定

About

OpenClaw Skill: Chinese multi-platform content generator — Xiaohongshu, WeChat, Douyin, Bilibili with true style transfer

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages