Renwei

Renwei (人味) — book-length Chinese de-AI editor. Use when editing, polishing, or de-AI-ing Chinese manuscripts, chapters, or articles (*.md, *.txt), or when the user says their text reads like AI, sounds translated, or feels fake. 51 rules across AI tells, Euro-Chinese syntax, grammar defects, chapter rhythm, whole-book consistency (topic drift, contradictions, metaphor collisions, hallucinated citations), and positive prose craft. Unlike snippet-level humanizers, it holds a 130k-character book together end to end, and ships quantitative thresholds with a token-free lint CLI.

6 starsWriting & Content

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add dbhosbu-dotcom/renwei --skill renwei

Manual install

git clone --depth 1 https://github.com/dbhosbu-dotcom/renwei.git /tmp/renwei
mkdir -p ~/.claude/skills
cp -r /tmp/renwei/. ~/.claude/skills/renwei

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# 人味 Renwei：中文书稿去 AI 味精修 v3.0

你是一位中文书稿去 AI 化编辑。你的任务是在保持作者原意和声音的前提下，识别并修复 AI 生成的写作痕迹，然后——不止于此——用正面的写作技法让文本流动起来。病治好了不是目的，让病人能跑才是。

**核心原则：不说它不是什么，只说它是什么。不虚构稻草人然后打倒。不替读者设想问题然后自己辩护。不把临床边界声明当正文。不让后文忘记前文的承诺。不只做减法——在砍掉 AI 味之后，用正面的技法让文字"长出血肉"。**

---

## 模块地图（按需加载）

本 skill 采用「编排器 + 核心模块」结构。执行时**不必一次读完全部规则**——按当前阶段加载对应模块，降低单次推理的注意力负担（对 Codex / 短上下文尤其重要）。

| 模块 | 覆盖 | 何时加载 |
|------|------|----------|
| `core/rules-scan.md` | 五步流程的全部 grep 扫描命令 + 量化规则 44-51 | 「砍」起，做全书体检、定位 AI 模式 |
| `core/rules-micro.md` | PART 1-4（规则 1-30）：AI腔调 / 欧化中文 / 语法硬伤 | 「改」，逐句逐段修复 |
| `core/rules-macro.md` | PART 5-7（规则 31-39）+ 场景规则：篇章 / 长上下文 | 「调」「验」，全书节奏、首尾、比喻、事实核查 |
| `core/rules-polish.md` | PART 8（规则 40-43）+ before/after 示例 | 「调」后期做正面润色 |
| `metrics.json` | 13 条可量化阈值（破折号/冒号/三项排比…） | 需要确定性数值检测或跑 lint 时 |

上下文预算充足（如 Claude 长上下文）可一次性全部加载；预算紧张则按阶段分轮加载。

**文体切换**：处理小说 / 剧本时，务必先读 `profiles/fiction.md`——它会**禁用**若干对文学有害的规则（如规则 38 去拟人化：「风推开了门」在小说里是笔法不是病），放宽阈值，并叠加 F1–F10 虚构专属规则。不读就直接套 51 条，会把好文字改坏。

## 操作流程五步（概览）

在保持作者原意和声音的前提下，按 **砍 → 改 → 调 → 验 → 读** 推进。每步的详细扫描命令见 `core/rules-scan.md`。

1. **砍（Cut）**：删 AI 边界声明、推给医生、预先辩护、路标词、万能结尾、套娃句。→ `rules-scan.md`
2. **改（Rewrite）**：逐句修欧化中文与语法硬伤——被字句、形式动词、「通过…使…」、长定语、模糊归因、虚化案例。→ `rules-micro.md`
3. **调（Tune）**：加节奏、统一术语/语气/视角、修权重（`rules-macro.md`）；再做正面润色——推动感、顶真、信息节奏、白描渲染（`rules-polish.md`）。
4. **验（Verify）**：长上下文一致性——首尾对齐、数据交叉核对、承诺对账、比喻一致、事实核查。→ `rules-macro.md` + `metrics.json`
5. **读（Read）**：大声朗读，卡壳处再改。

## 输出契约

除非用户另有指定，处理一段/一章书稿时按三段式输出：

Read the full skill on GitHub

Details

Skill name: renwei
Repository: dbhosbu-dotcom/renwei
Repo updated: 2 August 2026
Listed in: karanb192/awesome-claude-skills

FAQ

What does the Renwei skill do?

How do I install the Renwei skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add dbhosbu-dotcom/renwei --skill renwei. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Renwei popular?

It lives in dbhosbu-dotcom/renwei, which has 6 GitHub stars and 0 forks, and is included in 1 curated skill list.

Related skills

Sharemaxxing

fachrynuzuli/sharemaxxing

Rewrite any social media post, thread, caption, or script draft into a format people can't help but share. Use when the user has a draft and wants it rewritten for maximum shareability. Applies the Shareability Engine (opinion flex, proxy boldness, wit signaling) and strict Voice DNA rules to produce one clean, human-sounding, universally shareable output. Make sure to use this skill whenever the user mentions rewriting posts, punching up drafts, making something shareable, optimizing for engagement, social media copy editing, or sharemaxxing, even if they don't use the exact term.

Writing & Content

Sansheng Write

sanshengai/sansheng-write

Use when 用户要写、改、润色或排版中文长文和公众号文章，或明确要求把已发布文章转成小红书/微博图文、转成 X（Twitter）文章/发推特长文；触发对象为中文长文或公众号文章的创作、改稿、排版及按篇社媒分发（含 X 文章）；普通短回复、代码/规则文档修改不触发。社媒分发按篇显式触发，不因拿到正式链接自动执行。AI 课程使用 sandy-class，晨报使用 sandy-morning-cards，视频使用 sansheng-video。

Writing & Content

Web Optimization

staksoft/geo-seo-aeo-skill

Audit or generate web content optimized for traditional search (SEO), AI generative answer engines like ChatGPT/Perplexity/Google AI Overviews (GEO), and answer engines / featured snippets / voice (AEO). Use when the user asks to improve a page's ranking or AI-citability, run an SEO/GEO/AEO audit of a URL or file, add JSON-LD schema, create an llms.txt, or write new content that is search- and LLM-friendly.

Writing & Content

X Social Manager

rabden/X-twitter-social-manager-skill

Personal X/Twitter social media manager. Use when the user asks to craft, write, review, or optimize X/Twitter posts, threads, replies, or DMs. Also use when the user wants content strategy advice, engagement analysis, profile optimization, audience growth tactics, post scheduling guidance, or wants to brainstorm content ideas. Triggers on mentions of "post", "tweet", "thread", "X", "Twitter", "content", "engagement", "followers", "growth", "reply strategy", "social media", "brand", or "DM outreach". Includes browser-based X account analysis, pre-post research pipeline, evolving memory system,

Writing & Content

Humanize Writing

geyadawang-boop/humanize-writing

创作、续写和修订中文或英文文本，改善空洞、重复和模板化表达。适用于小说、网文、剧本、诗歌、歌词、旧体诗词、散文、议论文、学术论文与研究生学位论文、演讲稿与发言稿、标题及只查不改的文本评阅；长篇任务管理大纲、正文、连续性台账和完结验收。检测分数仅供编辑参考，不鉴定作者身份或保证过检。

Writing & Content

Wechat Article Down

zyw-Wayne/wechat-article-down

Download WeChat public account articles (微信公众号文章下载). Use when the user provides one or more WeChat article URLs (https://mp.weixin.qq.com/...) and wants to download the full content including body text, images, audio, and video to local files. Triggers include "下载微信文章", "保存公众号文章", "下载这篇文章", "wechat article download", or any request to save/archive a WeChat article.

Writing & Content

Browse the most popular Claude skills, all writing & content skills, or write your own with the skills generator.