According to AlphaMoat, as of 2026-09, AI Humanizer CN - 中文 AI 文本拟人优化 has 1,968 downloads and 3 stars — ranked #4,453 of 5,078 AI skills overall, and #298 of 362 in Content.
| Platform | Installs | Downloads | Stars | Comments |
|---|---|---|---|---|
| ClawHub ↗ | 42 | 1.5K | 3 | - |
| SkillHub ↗ | 214 | 2.0K | 3 | — |
Platforms count differently and update at different times.
The skill describes polishing Chinese content with different AI models, aiming to reduce awkward phrasing and improve readability. Results depend on the input, model, and configuration; the claim about removing AI traces is not a guarantee of passing detection.
Profile published 2026-10-27
Generated from the skill’s documentation as examples; not results of actual runs.
The counting method changed in 2026-09: skills with the same name are now counted separately by author, and the full catalog is covered. Dashed lines show months before the change; they are not compared with later months, and no growth is computed across the change.
| Month | Downloads | MoM | Stars |
|---|---|---|---|
| 2026-09 | 2.0K | — | 3 |
| 2026-08 | 1.5K | +10.0% | 3 |
| 2026-07 | 1.4K | +19.7% | 3 |
| 2026-06 | 1.2K | +10.6% | 3 |
| 2026-05 | 1.1K | — | 3 |
| 2026-03 | 636 | — | 3 |
The author’s original file, shown as plain text without edits or review by AlphaMoat. Links are not clickable; check commands before running them. Fetched 2026-10-23.
---
name: ai-humanizer-cn
description: 中文 AI 文本优化技能,支持多种 AI 模型(OpenAI/Anthropic/阿里),去除 AI 痕迹,保持专业性。
license: MIT
version: 5.0.0
author: pengong101
updated: 2026-03-18
metadata:
requires:
api_keys: - OPENAI_API_KEY
- ANTHROPIC_API_KEY
- DASHSCOPE_API_KEY
features:
- 多模型支持
- 风格调节
- 批量处理
- 代码优化
---
# AI Humanizer CN v5.0.0
**版本:** 5.0.0
**更新日期:** 2026-03-18
**作者:** pengong101
**许可:** MIT
---
## 🎯 核心功能
### 1. 多模型支持
**支持的 AI 模型:**
- ✅ OpenAI (GPT-4/GPT-3.5)
- ✅ Anthropic (Claude-3)
- ✅ 阿里百炼 (Qwen-Max/Qwen-Plus)
- ✅ MiniMax (ABAB 系列)
**模型选择:**
```bash
# 环境变量设置
export HUMANIZER_MODEL="openai/gpt-4"
# 或
export HUMANIZER_MODEL="anthropic/claude-3"
# 或
export HUMANIZER_MODEL="dashscope/qwen-max"
```
### 2. 风格调节
**支持的风格:**
- `formal` - 正式风格(技术文章、报告)
- `neutral` - 中性风格(新闻、说明)
- `casual` - 轻松风格(博客、社交媒体)
- `academic` - 学术风格(论文、研究)
- `business` - 商务风格(邮件、方案)
### 3. 批量处理
**支持的批量操作:**
- 文件夹批量处理
- 多格式混合(.md/.txt/.py/.js)
- 并行处理(多线程)
- 进度条显示
### 4. 代码优化
**支持的编程语言:**
- Python 代码注释优化
- JavaScript 代码注释优化
- SQL 查询语句优化
- Shell 脚本优化
---
## 💻 使用方式
### 方式 1:命令行调用
```bash
# 基础使用
humanizer-cn --input input.txt --output output.txt
# 指定风格
humanizer-cn --input input.txt --style formal
# 批量处理
humanizer-cn --batch ./docs --output ./output --style formal
# 代码优化
humanizer-cn --code input.py --output output.py
# 指定模型
humanizer-cn --input input.txt --model anthropic/claude-3
```
### 方式 2:Python 调用
```python
from ai_humanizer import Humanizer
# 初始化
h = Humanizer(
model="openai/gpt-4",
style="formal",
api_key="your_api_key"
)
# 优化文本
text = "本文旨在探讨..."
optimized = h.humanize(text)
print(optimized)
# 批量处理
results = h.batch_process(
input_dir="./docs",
output_dir="./output",
style="formal"
)
# 代码优化
code = "def foo(a,b):return a+b"
optimized_code = h.optimize_code(code, language="python")
print(optimized_code)
```
### 方式 3:OpenClaw 技能调用
```python
from skills.ai_humanizer_cn import humanize
result = humanize("AI 生成的文本", style="formal")
```
---
## ⚙️ 配置选项
### 环境变量
```bash
# API 密钥(至少配置一个)
export OPENAI_API_KEY="sk-xxx"
export ANTHROPIC_API_KEY="sk-ant-xxx"
export DASHSCOPE_API_KEY="sk-xxx"
# 默认模型
export HUMANIZER_MODEL="openai/gpt-4"
# 默认风格
export HUMANIZER_STYLE="formal"
# 批处理线程数
export HUMANIZER_THREADS="4"
# 日志级别
export HUMANIZER_LOG_LEVEL="INFO"
```
### 配置文件
**位置:** `~/.ai-humanizer/config.json`
```json
{
"model": "openai/gpt-4",
"style": "formal",
"max_length": 2000,
"temperature": 0.7,
"batch_size": 10,
"threads": 4,
"output_format": "text"
}
```
---
## 📊 性能指标
| 指标 | 数值 |
|------|------|
| 优化准确率 | 95%+ |
| 处理速度 | <500ms/千字 |
| 支持语言 | 中文、英文 |
| 批量处理 | 支持 1000+ 文件 |
| 并发处理 | 支持 10 线程 |
| 模型支持 | 4 家提供商 |
---
## 🧪 测试
### 运行测试
```bash
# 安装测试依赖
pip install pytest pytest-cov
# 运行测试
pytest tests/ -v --cov=ai_humanizer
# 查看覆盖率
coverage html
```
### 测试覆盖
```
Name Stmts Miss Cover
-------------------------------------------
ai_humanizer.py 250 25 90%
humanize_v5.py 180 18 90%
tests/test_humanizer.py 150 0 100%
-------------------------------------------
TOTAL 580 43 93%
```
---
## 📦 文件结构
```
ai-humanizer-cn/
├── SKILL.md # 技能文档(本文件)
├── README.md # 详细说明
├── LICENSE # MIT 许可证
├── clawhub.json # ClawHub 配置
├── requirements.txt # Python 依赖
├── setup.py # 安装脚本
├── ai_humanizer.py # 主程序(v5.0.0)
├── humanize_v5.py # v5.0.0 核心
├── humanize_v4.py # v4.0.0(兼容)
├── humanize_v3.py # v3.0.0(兼容)
├── config.py # 配置管理
├── utils.py # 工具函数
├── tests/ # 测试目录
│ ├── test_humanizer.py
│ └── fixtures/
├── examples/ # 示例目录
│ ├── basic_usage.py
│ └── batch_processing.py
└── docs/ # 文档目录
├── installation.md
├── usage.md
└── api.md
```
---
## 🔧 安装
### 方式 1:pip 安装
```bash
pip install ai-humanizer-cn
```
### 方式 2:源码安装
```bash
git clone https://github.com/pengong101/ai-humanizer-cn
cd ai-humanizer-cn
pip install -e .
```
### 方式 3:ClawHub 安装
```bash
openclaw skills install ai-humanizer-cn
```
---
## 📊 版本历史
| 版本 | 日期 | 主要更新 |
|------|------|---------|
| **v5.0.0** | 2026-03-18 | 多模型支持/批量处理/代码优化/测试覆盖 |
| v4.0.0 | 2026-03-18 | 代码优化/表格格式化/PPT 大纲 |
| v3.1.0 | 2026-03-17 | 8 维风格向量/多语言支持 |
| v3.0.0 | 2026-03-14 | 7 种写作风格/语境感知 |
| v2.1.0 | 2026-03-13 | 批量处理/性能优化 |
| v2.0.0 | 2026-03-12 | 架构重构/风格调节 |
| v1.1.0 | 2026-03-11 | 性能优化 |
| v1.0.0 | 2026-03-11 | 初始版本 |
---
## 🔗 相关链接
- **GitHub:** https://github.com/pengong101/ai-humanizer-cn
- **PyPI:** https://pypi.org/project/ai-humanizer-cn/
- **ClawHub:** 待发布
- **文档:** https://ai-humanizer-cn.readthedocs.io/
- **作者:** pengong101
---
## 📝 常见问题
### Q: 需要 API 密钥吗?
**A:** 是的,至少需要配置一个 AI 模型的 API 密钥:
- OpenAI: `OPENAI_API_KEY`
- Anthropic: `ANTHROPIC_API_KEY`
- 阿里百炼:`DASHSCOPE_API_KEY`
### Q: 支持哪些文件格式?
**A:** 支持:
- 文本:.txt, .md
- 代码:.py, .js, .sql, .sh
- 批量:文件夹递归处理
### Q: 批量处理速度如何?
**A:** 默认 4 线程并行,每秒可处理 10-20 个文件(取决于文本长度)。
---
**最后更新:** 2026-03-18
**版本:** 5.0.0 (Latest)
**许可:** MIT License
**测试覆盖:** 93%Detects and removes AI writing traces so text reads more like a person wrote it.
A writing-editor skill that finds the usual marks of AI-generated text and rewrites them into more natural prose without losing the meaning.
A Chinese copy-editing skill that spots and removes the traces of AI-generated text.
wechat-search finds popular WeChat articles and organizes results by relevance, popularity and recency.
Extracts titles, descriptions, and spoken scripts from public short-video share links, with rewritten versions.
humanizer guides natural expression and style editing for Chinese and English text.
Writes web-novel chapters into files and replies in chat with summaries and pointers only.
Bundle a set of marketing playbooks into one skill and route each request to the module that fits.
Data month: 2026-09 · Downloads, stars and installs are aggregated monthly from public skill registries (ClawHub, SkillHub). See methodology.