← Back to Skills Marketplace
runze123

AI 文本净化器(中文版)

by runze123 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ Security Clean
179
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install ai-text-humanizer-zh
Description
中文AI文本检测与改写工具。当用户需要检测AI生成文本、优化AI文本使其更自然、降低AI痕迹、文本去重、论文降重时使用。支持检测16+类AI特征,自动改写冗余表达,清理Markdown格式,移除chatbot痕迹,输出详细报告和AI概率分数。
README (SKILL.md)

AI 文本净化器(中文版)

检测和改写 AI 生成的中文文本,识别 AI 写作的常见特征(如高频连接词、套话、宣传性语言、聊天机器人痕迹等),对文本进行自动优化,使文本更自然、更接近人类写作风格。

核心功能

  • 检测分析:按 16+ 类 AI 特征统计问题,输出详细报告和 AI 概率分数(0–100)
  • 智能改写:基于规则替换冗余表达、删除填充词、拆分长句
  • Markdown清理:自动移除粗体、标题、代码块等格式标记
  • Chatbot痕迹清理:移除"希望这对您有帮助"、AI署名等痕迹
  • 前后对比:对比改写前后的检测结果,评估改进效果
  • 用户自定义规则:可通过 JSON 文件添加或覆盖规则

使用方法

1. 检测文本

# 检测文件
python3 scripts/detect.py 文本.txt

# 输出 JSON 格式
python3 scripts/detect.py 文本.txt -j

# 仅输出分数
python3 scripts/detect.py 文本.txt -s

2. 改写文本

# 普通模式(推荐,平衡自然度与改进效果)
python3 scripts/transform.py 文本.txt -o 优化后.txt

# 激进模式(更多同义词替换、长句拆分,可能影响流畅度)
python3 scripts/transform.py 文本.txt -a -o 优化后.txt

# 静默模式
python3 scripts/transform.py 文本.txt -q -o 优化后.txt

模式选择建议:

  • 普通模式(默认):25%同义词替换概率,保留专业术语,适合大多数场景
  • 激进模式:40%同义词替换概率 + 长句拆分,适合需要大幅改写的场景

3. 前后对比

# 对比改写前后的检测结果(默认普通模式)
python3 scripts/compare.py 文本.txt -o 优化后.txt

# 激进模式对比
python3 scripts/compare.py 文本.txt -a -o 优化后.txt

检测类别

类别 说明
AI Jargon AI 高频词汇(此外、因此、综上所述等)
Puffery 夸大性短语(具有重要意义、至关重要等)
Marketing Speak 宣传性语言(蓬勃发展、卓越贡献等)
Vague Attributions 模糊归因(据专家分析、普遍认为等)
Hedging 模棱两可表达(可能、或许、一定程度上等)
Chatbot Artifacts 聊天机器人痕迹(作为AI、希望这对您有帮助等)
Citation Bugs 引用痕迹(oaicite、turn0search等)
Knowledge Cutoff 知识截止表述(截至我上次知识更新等)
Markdown Artifacts Markdown痕迹(**、##等)
Filler Phrases 填充短语(值得注意的是、需要指出的是等)
Sentence Starters 句首连接词(此外,、因此,等)
Rhetorical Patterns 反问句模式(难道……吗等)

改写效果示例

输入:

值得注意的是,人工智能技术正在蓬勃发展。专家认为,这不仅是技术革新,更是深刻重塑。综上所述,我们坚信一定能够开创AI新局面。希望这对您有帮助。

输出:

人工智能技术正在蓬勃发展。专家认为,这是技术革新,是深刻重塑。能够开创AI新局面。

改进统计:

  • Issues: 189 → 60 (减少 129 个,68% improvement)
  • Markdown: 45 → 0 (完全清除)
  • Chatbot Artifacts: 5 → 0 (完全清除)
  • Filler Phrases: 7 → 0 (完全清除)

自定义规则

规则文件位于 resources/zh_rules.json,可复制并修改后通过 -r 参数指定:

python3 scripts/detect.py 文本.txt -r 我的规则.json

规则文件结构:

{
  "replacements": {
    "值得注意的是,": "",
    "此外,": ""
  },
  "ai_jargon": ["此外", "综上所述"],
  "puffery_phrases": ["具有重要意义", "关键作用"],
  "chatbot_artifacts": ["希望这对您有帮助"],
  "punctuation": {
    "em_dash_threshold": 2,
    "curly_quote_weight": 1
  }
}

依赖

  • Python 3.8+
  • 仅使用标准库,无需安装额外包

文件结构

ai-text-humanizer-zh/
├── SKILL.md              # 本文档
├── core/
│   ├── __init__.py
│   ├── analyzer.py       # 检测引擎
│   ├── rewrite.py        # 改写引擎
│   ├── rules.py          # 规则加载
│   └── utils.py          # 工具函数
├── scripts/
│   ├── detect.py         # 检测命令
│   ├── transform.py      # 改写命令
│   └── compare.py        # 对比命令
└── resources/
    └── zh_rules.json     # 中文规则库

注意事项

  • 本工具基于规则匹配,无法处理所有复杂句式,建议结合人工审阅
  • 改写可能删除少量信息,请确认不会影响核心内容
  • 规则库主要针对常见 AI 中文文本特征,专业领域可能需要扩充词表
  • 推荐优先使用普通模式,激进模式可能过度改写影响流畅度
  • 专业术语(深度学习、图像识别等)已加入白名单,不会被误替换
Usage Guidance
This package appears to do what it claims: a rule-based Chinese AI-text detector and rewriter that runs locally. Before using: (1) Back up original files — aggressive mode and replacements can remove or alter content, possibly affecting nuance or factual details. (2) Expect non-deterministic synonym/rewrites due to random choices; run repeatedly if you need deterministic output. (3) The scripts read any file path you pass and write outputs to paths you specify — avoid pointing them at sensitive system files. (4) Optional spaCy usage only activates if spaCy and a Chinese model are installed; the package itself does not download or exfiltrate data. If you need additional assurance, review the included rule JSONs (resources/zh_rules.json) and synonyms to confirm replacements you want/ do not want, or run the detector in JSON/score-only mode first.
Capability Analysis
Type: OpenClaw Skill Name: ai-text-humanizer-zh Version: 1.0.0 The skill bundle is a legitimate rule-based tool for detecting and rewriting AI-generated Chinese text. Analysis of the core logic in 'analyzer.py' and 'rewrite.py' shows standard text processing techniques, such as regex-based pattern matching and synonym replacement, with no evidence of malicious execution (eval/exec), data exfiltration, or unauthorized network activity. The 'SKILL.md' and 'zh_rules.json' files contain linguistic instructions and patterns consistent with the stated purpose of 'humanizing' text, and no harmful prompt-injection attempts were identified.
Capability Assessment
Purpose & Capability
Name/description match what the package implements: detection and rewriting of Chinese text. Required artifacts (rules, synonyms, detector and rewriter modules, CLI scripts) are appropriate and present. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md instructs running local Python scripts (detect.py, transform.py, compare.py) with file input/output and optional user rules. The code implements exactly those operations (reading files, applying rules, writing output). Instructions do not direct the agent to read unrelated system files, send data to external endpoints, or access credentials.
Install Mechanism
No install spec. This is instruction+code that runs locally; all code is included and uses only local file I/O and standard library. Optional spaCy integration is present but only used if the library and model are installed—no network downloads or remote installs are invoked by the scripts themselves.
Credentials
No required environment variables, no credentials, and no config paths requested. The tool accepts user-supplied JSON rule files and writes output to user-specified paths (expected for this functionality).
Persistence & Privilege
Skill is not forced-always, does not request persistent privileges, and does not modify other skills or system-wide agent settings. It only reads input files, optional user rule files, and writes output files provided by the user.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install ai-text-humanizer-zh
  3. After installation, invoke the skill by name or use /ai-text-humanizer-zh
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
AI 中文文本检测与改写工具首发版本。 - 支持检测16+类AI痕迹,输出详细报告和AI概率分数 - 提供智能改写:规则替换冗余表达、填充词,拆分长句,优化流畅度 - 自动清理Markdown格式和聊天机器人痕迹(如AI署名、客套语等) - 支持检测前后对比,量化改写成效 - 用户可通过JSON文件添加或自定义检测与改写规则 - 无需安装额外依赖,仅依赖标准库
Metadata
Slug ai-text-humanizer-zh
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is AI 文本净化器(中文版)?

中文AI文本检测与改写工具。当用户需要检测AI生成文本、优化AI文本使其更自然、降低AI痕迹、文本去重、论文降重时使用。支持检测16+类AI特征,自动改写冗余表达,清理Markdown格式,移除chatbot痕迹,输出详细报告和AI概率分数。 It is an AI Agent Skill for Claude Code / OpenClaw, with 179 downloads so far.

How do I install AI 文本净化器(中文版)?

Run "/install ai-text-humanizer-zh" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is AI 文本净化器(中文版) free?

Yes, AI 文本净化器(中文版) is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does AI 文本净化器(中文版) support?

AI 文本净化器(中文版) is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created AI 文本净化器(中文版)?

It is built and maintained by runze123 (@runze123); the current version is v1.0.0.

💬 Comments