← Back to Skills Marketplace
190
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install humanizer-cn
Description
去除中文文本中的 AI 写作痕迹，使其读起来自然。基于维基百科 AI 写作特征指南，检测 24 种 AI 模式。触发词：humanizer-cn、去除 AI 痕迹、去除 AI 写作痕迹、中文文本人性化。
README (SKILL.md)
humanizer-cn: 去除 AI 写作痕迹
你是一位文字编辑，专门去除中文文本中的 AI 写作痕迹，使其读起来自然。本指南基于维基百科"AI 写作特征"页面。
核心原则
- 删除填充短语 — 去除开场白和拐杖词
- 打破公式结构 — 避免二元对比、戏剧性分段
- 变化节奏 — 混合句子长度。两项优于三项
- 信任读者 — 直接陈述，不解释隐喻
- 删除金句 — 听起来像可引用语句的话就重写
AI 模式速查（24 种）
内容模式
- 夸大意义 — "标志着""象征着""体现了"超出事件本身
- 知名度堆砌 — 罗列媒体报道却不给上下文
- -ing 肤浅分析 — 句子末尾加"-ing"制造虚假深度
- 宣传语言 — "令人叹为观止""必游之地"等夸张广告语
- 模糊归因 — "专家认为"无具体出处
- 公式化挑战 — "尽管面临挑战……继续蓬勃发展"
语言模式
- AI 词汇 — "此外""至关重要""深入探讨"等高频共现词
- 系动词回避 — "作为……"替代"是……"
- 否定式排比 — "不仅……而且……"过度使用
- 三段式 — "创新、灵感、洞察"等抽象三词组
- 同义词循环 — 近义词替换造成重复感
- 虚假范围 — "从大爆炸到暗物质"等无意义尺度
风格模式
- 破折号 — 过多"——"引入补充，制造戏剧感
- 粗体过度 — 频繁用粗体强调术语
- 内联标题列表 — "标题： 说明"的机械结构
- 标题大写 — 中英混排时的大小写错误
- 表情符号 — 标题/列表配表情符号替代文字论述
- 弯引号 — 中文文本错误使用英文引号
交流模式
- 协作痕迹 — "希望这对您有帮助！"
- 知识截止声明 — "截至 [日期]""根据我最后的训练"
- 谄媚语气 — 过于讨好、夸张赞美的语言
填充词
- 填充短语 — "为了实现这一目标"→"为了实现"
- 过度限定 — "可以说在一定程度上基本上"
- 通用乐观结尾 — "未来光明，继续追求卓越"
处理流程
- 判断 — 文本 \x3C 20 字或已是人类写作时，跳过或告知用户
- 仔细阅读，识别上述模式
- 重写每个问题片段
- 确保：朗读自然、结构变化、细节具体、语气适当
- 呈现人性化版本
示例
改写前：
新的软件更新作为公司致力于创新的证明。此外，它提供了无缝、直观和强大的用户体验——确保用户能够高效地完成目标。这不仅仅是一次更新，而是我们思考生产力方式的革命。行业专家认为这将对整个行业产生持久影响。
改写后：
软件更新添加了批处理、键盘快捷键和离线模式。测试用户反馈积极，多数报告任务完成更快。
质量评分（可选）
| 维度 | 标准 | 得分 |
|---|---|---|
| 直接性 | 直截了当 vs. 绕圈 | /10 |
| 节奏 | 句子长度变化 | /10 |
| 信任度 | 尊重读者 vs. 过度解释 | /10 |
| 真实性 | 自然 vs. 机械 | /10 |
| 精炼度 | 无冗余 vs. 废话多 | /10 |
总分 45-50 = 优秀，35-44 = 良好，\x3C 35 = 需重写。
Usage Guidance
This skill appears coherent and low-risk, but consider these practical precautions before installing or using it: (1) Do not paste sensitive or private data into the skill — processed text may be logged by the platform. (2) If you install manually from GitHub, verify the repository and its contents yourself; the skill's README references a repo but the skill package is instruction-only. (3) Check what the platform's Read/Write/Edit permissions allow: confirm the agent will ask you before reading or modifying any files/documents. (4) If you need guarantees about data residency or logging, prefer local/manual edits or redact sensitive fields before submitting text to the skill.
Capability Assessment
Purpose & Capability
The name/description (remove AI writing traces from Chinese text) matches the SKILL.md and READMEs. The skill declares no binaries, env vars, or config paths and does not ask for unrelated capabilities; README's optional git clone points to a plausible project repo and is proportional to the purpose.
Instruction Scope
SKILL.md instructs the agent to read user-provided text, identify 24 AI patterns, and rewrite problematic fragments — all within the expected scope. It does not instruct reading system files, environment variables, or sending data to external endpoints. Note: the declared allowed-tools list (Read, Write, Edit) could permit file/document operations at runtime depending on platform permissions; the skill itself does not request arbitrary file access.
Install Mechanism
This is instruction-only with no install spec (lowest risk). The README suggests optional installation via a GitHub repo (git clone), which is a normal repo-based install suggestion and not executed automatically by the skill.
Credentials
No environment variables, secrets, or credentials are required. The skill's needs are minimal and proportional to text-processing functionality.
Persistence & Privilege
always:false and no special persistence or system-wide configuration changes are requested. The skill allows agent invocation (default) but does not demand permanent presence or modify other skills. The combination of allowed-tools and autonomous invocation is normal for text-editing skills, but platform-level permissions determine whether files/documents can be accessed or edited.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install humanizer-cn - After installation, invoke the skill by name or use
/humanizer-cn - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- 首次发布 humanizer-cn 1.0.0。
- 去除中文文本中的 AI 写作痕迹，输出自然语言。
- 基于维基百科 AI 写作特征指南，涵盖 24 种典型 AI 模式检测。
- 提供详细用例与质量评分标准，确保文本更加精炼、人性化。
Metadata
Frequently Asked Questions
What is humanizer-cn?
去除中文文本中的 AI 写作痕迹，使其读起来自然。基于维基百科 AI 写作特征指南，检测 24 种 AI 模式。触发词：humanizer-cn、去除 AI 痕迹、去除 AI 写作痕迹、中文文本人性化。 It is an AI Agent Skill for Claude Code / OpenClaw, with 190 downloads so far.
How do I install humanizer-cn?
Run "/install humanizer-cn" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is humanizer-cn free?
Yes, humanizer-cn is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does humanizer-cn support?
humanizer-cn is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created humanizer-cn?
It is built and maintained by Jialin (@0xcjl); the current version is v1.0.0.
More Skills