<!-- WEHUB_ZH_README -->
> [!NOTE]
> 本文档由 WeHub 基于上游 README 翻译整理，属于社区翻译，非官方中文文档。  
> [English](./README.en.md) · [原始项目](https://github.com/blader/humanizer) · [上游 README](https://github.com/blader/humanizer/blob/HEAD/README.md)  
> 原作者、版权与许可证归属以原始项目及本仓库 LICENSE 文件为准。

# Humanizer

一个可移植的 agent skill，用于去除文本中的 AI 生成痕迹，使其听起来更自然、更像人写的。它就是纯 Markdown，因此可在任何支持 skill 式指令的运行环境中使用。

## Installation

### Skills CLI

使用跨 agent 的 skills CLI 安装：

```bash
npx skills add blader/humanizer
```

更新已有安装：

```bash
npx skills update humanizer
```

安装到所有受支持的 agent 运行环境：

```bash
npx skills add blader/humanizer --agent '*'
```

若只针对某个已配置的运行环境，传入其 agent 名称：

```bash
npx skills add blader/humanizer --agent <agent-name>
```

### Claude Code plugin

Claude Code 用户也可以将 Humanizer 作为插件安装：

```
/plugin marketplace add blader/humanizer
/plugin install humanizer@humanizer
```

之后可通过 `/humanizer:humanizer` 调用该 skill。

### Manual

任何 agent 运行环境都可以直接使用该 skill，因为运行时产物就是 `SKILL.md`。将其安装到你的运行环境所期望的 skill 目录中，或把 `SKILL.md` 复制进已有的 skill 文件夹。

例如：

```bash
git clone https://github.com/blader/humanizer.git /path/to/your/skills/humanizer
```

或者，如果你已经克隆了本仓库：

```bash
mkdir -p /path/to/your/skills/humanizer
cp SKILL.md /path/to/your/skills/humanizer/
```

## Usage

按你的 agent 运行环境暴露已安装 skill 的方式来调用。常见形式包括斜杠命令或直接请求：

```
/humanizer

[paste your text here]
```

```
Please humanize this text: [your text]
```

### Voice Calibration

若要贴合你的个人写作风格，请提供一段你自己的写作样本：

```
/humanizer

Here's a sample of my writing for voice matching:
[paste 2-3 paragraphs of your own writing]

Now humanize this text:
[paste AI text to humanize]
```

该 skill 会分析你的句式节奏、用词与个人习惯，并将其应用到改写中，而不是产出泛泛的“干净”文本。

## Overview

基于 [Wikipedia 的 “Signs of AI writing”](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) 指南，由 WikiProject AI Cleanup 维护。这份全面指南来自对数千例 AI 生成文本的观察。

该 skill 还包含最后一轮“明显由 AI 生成”的审计，以及第二次改写，以抓出初稿中残留的 AI 痕迹。

### Key Insight from Wikipedia

> “LLMs 使用统计算法猜测接下来应该出现什么。结果往往趋向于在最广泛情形下最统计上可能的答案。”

## 33 Patterns Detected (with Before/After Examples)

### Content Patterns

| # | Pattern | Before | After |
|---|---------|--------|-------|
| 1 | **Significance inflation（意义夸大）** | "marking a pivotal moment in the evolution of..." | "was established in 1989 to collect regional statistics" |
| 2 | **Notability name-dropping（名气堆砌）** | "cited in NYT, BBC, FT, and The Hindu" | "In a 2024 NYT interview, she argued..." |
| 3 | **Superficial -ing analyses（浅层 -ing 式分析）** | "symbolizing... reflecting... showcasing..." | 删除，或用真实来源展开 |
| 4 | **Promotional language（宣传腔）** | "nestled within the breathtaking region" | "is a town in the Gonder region" |
| 5 | **Vague attributions（含糊归因）** | "Experts believe it plays a crucial role" | "according to a 2019 survey by..." |
| 6 | **Formulaic challenges（公式化挑战）** | "Despite challenges... continues to thrive" | 写出关于实际挑战的具体事实 |

### Language Patterns

| # | Pattern | Before | After |
|---|---------|--------|-------|
| 7 | **AI vocabulary（AI 词汇）** | "Actually... additionally... testament... landscape... showcasing" | "also... remain common" |
| 8 | **Copula avoidance（回避系动词）** | "serves as... features... boasts" | "is... has" |
| 9 | **Negative parallelisms / tailing negations（否定平行结构 / 尾部否定）** | "It's not just X, it's Y", "..., no guessing" | 直接陈述要点 |
| 10 | **Rule of three（三项法则）** | "innovation, inspiration, and insights" | 使用自然数量的条目 |
| 11 | **Synonym cycling（同义词轮换）** | "protagonist... main character... central figure... hero" | "protagonist"（在最清晰时重复使用） |
| 12 | **False ranges（虚假范围）** | "from the Big Bang to dark matter" | 直接列出主题 |
| 13 | **Passive voice / subjectless fragments（被动语态 / 无主语片段）** | "No configuration file needed" | 在有助于清晰时点明行为主体 |

### Style Patterns

| # | Pattern | Before | After |
|---|---------|--------|-------|
| 14 | **Em/en dashes（破折号）** | "institutions—not the people—yet this continues—" | 删去：改用句号、逗号、冒号或括号 |
| 15 | **Boldface overuse（过度加粗）** | "**OKRs**, **KPIs**, **BMC**" | "OKRs, KPIs, BMC" |
| 16 | **Inline-header lists（行内标题列表）** | "**Performance:** Performance improved" | 改写成正文 |
| 17 | **Title Case Headings（标题式大小写）** | "Strategic Negotiations And Partnerships" | "Strategic negotiations and partnerships" |
| 18 | **Emojis** | "🚀 Launch Phase: 💡 Key Insight:" | 删除 emoji |
| 19 | **Curly quotes（弯引号）** | `said “the project”` | `said “the project”` |
| 26 | **Hyphenated word pairs（连字符词对）** | “cross-functional, data-driven, client-facing” | 常见词对去掉连字符 |
| 27 | **Persuasive authority tropes（权威说服套话）** | "At its core, what matters is..." | 直接陈述要点 |
| 28 | **Signposting announcements（路标式开场）** | "Let's dive in", "Here's what you need to know" | 直接进入内容 |
| 29 | **Fragmented headers（割裂式标题）** | "## Performance" + "Speed matters." | 让标题本身完成工作 |
| 30 | **Diff-anchored writing（以 diff 为锚的写法）** | "This function was added to replace..." | 描述它做什么，而不是改了什么 |
| 31 | **Manufactured punchlines / staccato drama（人造金句 / 断奏式戏剧感）** | "It had no preference. No prior. No nostalgia." | 使用多变的句长与具体主张 |
| 32 | **Aphorism formulas（格言公式）** | "Symmetry is the language of trust" | 用实际主张替换公式 |
| 33 | **Conversational rhetorical openers（对话式修辞开场）** | "Honestly? It depends..." | 去掉假装坦诚的铺垫 |

### Communication Patterns

| # | Pattern | Before | After |
|---|---------|--------|-------|
| 20 | **Chatbot artifacts（聊天机器人痕迹）** | "I hope this helps! Let me know if..." | 整段删除 |
| 21 | **Cutoff disclaimers（信息不足免责声明）** | "While details are limited in available sources..." | 找到来源，或删除 |
| 22 | **Sycophantic tone（谄媚语气）** | "Great question! You're absolutely right!" | 直接回应 |

### Filler and Hedging

| # | Pattern | Before | After |
|---|---------|--------|-------|
| 23 | **Filler phrases（填充短语）** | "In order to", "Due to the fact that" | "To", "Because" |
| 24 | **Excessive hedging（过度对冲）** | "could potentially possibly" | "may" |
| 25 | **Generic conclusions（空泛结论）** | "The future looks bright" | 具体计划或事实 |

## Full Example

**Before（听起来像 AI）：**
> I recently spent five unforgettable days in Lisbon, and let me tell you — this city completely stole my heart. From the moment I arrived, I knew I was somewhere truly special.
>
> Nestled along the banks of the Tagus River, Lisbon stands as a vibrant testament to Portugal's enduring spirit, where rich history and modern energy intertwine at every turn. Yes, the famous hills are challenging — my legs certainly felt it! — but every climb rewards you with breathtaking, panoramic views that make it all worthwhile.
>
> No trip would be complete without riding the iconic Tram 28, winding through the city's most historic neighborhoods. And the food? Simply divine. The original pastéis de nata at Pastéis de Belém are a beloved national treasure, and savoring one still warm was a moment I will never forget.
>
> But what truly makes Lisbon special isn't just the sights — it's the feeling. Wander a few steps off the main squares and you'll discover a quieter, more authentic side: sun-drenched alleys, charming tiled facades, and friendly locals going about their daily lives, inviting you to slow down and savor every moment.
>
> Of course, the must-see São Jorge Castle offers stunning views over the rooftops below. The queues can be long, but the experience is absolutely worth it for any history buff or curious traveler.
>
> Would I go back? Absolutely. Lisbon isn't just a place to visit — it's a place to fall in love with, again and again. If you're dreaming of your next getaway, this is one destination that promises memories to last a lifetime. ✨

**之后（人性化版本）：**
> 去年十月我在里斯本待了五天，到现在心情还是五味杂陈。很美，没错。也比任何人提醒我的更费膝盖。
>
> 山丘才是整座城市的全部故事，却 somehow 从未出现在宣传册里。我住的酒店在 Alfama，拍照很好看，实际体验则是：每次想喝杯咖啡，都像在爬一段六层楼的楼梯。第二天小腿就开始抗议。我开始按哪条路下坡来规划每次步行——用这种方式看一座城市很奇怪，但也许很真实。
>
> 人人都说该坐 28 路电车，我也坐了，在陌生人的背包旁挤了四十分钟，同时有三个旅行团在拍摄同一个街角。下次我会步行走这条线，或者早餐前去。不过蛋挞（custard tarts）确实名不虚传。我在 Graça 一家不起眼的小店买了一个，还热乎着，大约三十秒里我明白了人们为什么会为糕点专门规划一趟旅行。
>
> 我没料到的是，离开主要广场后城市会那么安静。从任何广场往上走两个街区，就会变成晾衣绳、斑驳瓷砖、敞开的窗户，还有把电视音量开很大的老人在看足球。那才是我一直惦记的里斯本，不是城堡。
>
> 城堡还行。景色很好，队伍很长，我花在排队挪向入口上的时间，比进去后真正看东西的时间还多。如果只有两天，我会拿它换一下午随便乱逛。
>
> 我会再去，但要选春天，还要穿更好的鞋。里斯本不会刻意为你把一切都变得轻松。我想我喜欢这一点，即使我的腿不同意。

## 参考资料

- [Wikipedia: Signs of AI writing](https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing) - Primary source
- [WikiProject AI Cleanup](https://en.wikipedia.org/wiki/Wikipedia:WikiProject_AI_Cleanup) - Maintaining organization

## 版本历史

- **2.8.2** - 将完整的 before/after 示例替换为第一人称里斯本旅行回顾。after 版本在保持与 before 相同主题、视角和大致篇幅的同时，去除了 AI 写作痕迹，且没有变得生硬或口号化。33 条模式无变化。
- **2.8.1** - 新增跨 agent 安装文档、可选的 Claude Code 插件打包，以及紧凑的二手文本误报防护。33 条模式无变化。
- **2.8.0** - 新增风格/节奏模式 #31-33，针对刻意制造的 punchline、格言式公式和对话式修辞开场；扩展 #20 以捕获「主动提出继续对话」的 chatbot 式结尾。共 33 条模式。
- **2.7.0** - 新增模式 #30（diff-anchored writing）；将 em/en dash 从「过度使用」改为硬性剔除；扩展 #21 以覆盖推测性补全（如 "maintains a low profile"）。共 30 条模式。
- **2.6.0** - 清理整理：合并重复的工作流章节，将个性指导限定在需要表达声音的内容上，移除模型指纹识别子章节，并压缩示例。29 条模式无变化。
- **2.5.1** - 新增被动语态/无主语片段规则，总数提升至 29 条模式
- **2.5.0** - 新增说服性 framing、路标式 signposting 和碎片化标题的模式；扩展否定平行结构以覆盖尾部否定；收紧 em dash 过度使用的表述；修正 frontmatter 措辞为 "filler phrases"
- **2.4.0** - 新增声音校准：根据用户样本文本匹配其个人写作风格
- **2.3.0** - 新增模式 #25：连字符词对过度使用
- **2.2.0** - 新增最终「明显 AI 生成」审计 + 第二轮改写提示
- **2.1.1** - 修正模式 #18 示例（弯引号 vs 直引号）
- **2.1.0** - 为全部 24 条模式新增 before/after 示例
- **2.0.0** - 基于 Wikipedia 原文内容完整重写
- **1.0.0** - 初始发布

## 许可证

MIT