微信公众号发布工具

微信公众号文章发布工具，支持 Markdown 转换、AI 去痕、多主题

Security Scan

The skill's code and documentation mostly match its stated purpose (WeChat publishing + AI 'humanize'), but there are inconsistencies around declared requirements and an unexpected behavior (auto-loading ~/.openclaw/.env) that warrant caution before installing.

Purpose & Capability

The name/description (WeChat publisher: Markdown→HTML, AI de-identification, themes, publish to WeChat) aligns with the included code and CLI. However the registry metadata claims no required environment variables or config paths while the SKILL.md and code clearly expect WeChat credentials (AppID/AppSecret) and optional AI API keys—this mismatch is an incoherence in declared requirements.

Instruction Scope

Runtime instructions and the provided scripts/CLI are focused on conversion and publishing. However the bash helper (scripts/publish.sh) and Python config code will automatically load environment variables from ~/.openclaw/.env when present. That file is external to the skill and may contain unrelated secrets; auto-loading it widens the agent's data surface implicitly. The skill also supports fetching remote CSS and (optionally) using external conversion APIs or AI providers (expected), which will transmit article content to remote endpoints if enabled.

Install Mechanism

There is no formal install spec in the registry, but SKILL.md instructs pip install from a GitHub repo. Packaging appears standard (pyproject.toml). This is a normal install path; nothing like arbitrary binary downloads or opaque installers were present in the manifest. Verify the GitHub repo/source before installing.

Credentials

The skill legitimately needs WeChat AppID/AppSecret and (optionally) AI API credentials. But the registry metadata did not declare required env vars/config paths. More importantly, the code explicitly reads ~/.openclaw/.env (and the shell script will source it), which can import arbitrary environment variables from the OpenClaw environment into the process—potentially exposing unrelated tokens/config to the skill. The skill only reads specific keys from that file (WECHAT_*/AI_*), but the shell sourcing behavior exports all variables globally to the process.

Persistence & Privilege

The skill does not request 'always: true' or other elevated persistent privileges. It stores a config under ~/.wechat-publish-pro/ (normal) and does not appear to modify other skills or system-wide agent settings.

What to consider before installing

What to check before installing: 1) Verify the source: SKILL.md suggests pip installing from a GitHub repo—confirm the repository and commit integrity before installing. 2) Expect to provide sensitive credentials: you will need WeChat AppID/AppSecret and (optionally) an AI provider API key—store these only where you intend. 3) Beware ~/.openclaw/.env auto-loading: the provided shell script and config code will load ~/.openclaw/.env if present (the shell script 'source's it and exports all variables). That can expose unrelated secrets in that file to the publishing process. If you keep other tokens in ~/.openclaw/.env, remove them or run the tool in an isolated environment. 4) Review network behavior: the converter can optionally call external conversion APIs and the humanizer will call configured AI provider endpoints—if you enable those, article text and images may be transmitted off-host. 5) If you need higher assurance: inspect platforms/wechat.py to confirm it only talks to official WeChat endpoints and that no unexpected remote endpoints are hard-coded; ask the maintainer to update registry metadata to declare required env vars and to stop implicitly sourcing ~/.openclaw/.env. If the maintainer provides those clarifications or removes the auto-load behavior, this assessment could be upgraded to benign.

wechat-publish-pro Skill

微信公众号文章发布工具，完全独立实现，不依赖任何外部 CLI 工具。

功能特性

• 纯 Python 实现: 不依赖 wenyan-cli 或其他外部工具
• Markdown 转换: 内置 Markdown → HTML 转换，参考 wenyan-cli 排版风格
• 多主题支持: 多种精美主题（green、blue、purple、orange、default、simple）
• 代码高亮: Mac 风格代码块，语法高亮
• AI 去痕: 发布前自动 AI 去痕处理，让内容读起来更像真人写的
• 封面生成: 自动处理封面图，支持本地/网络图片
• 直接发布: 一键发布到微信公众号草稿箱

触发条件

当用户需要以下操作时自动触发：

• 将 Markdown 或 HTML 文章发布到微信公众号草稿箱
• 转换 Markdown 为公众号格式 HTML
• 使用特定主题发布公众号文章
• 测试微信连接

安装

# 从 GitHub 安装
pip install git+https://github.com/yuesf/wechat-publish-pro.git

配置

第一步：获取微信公众号凭证

1. 登录 https://developers.weixin.qq.com/ 公众号平台
2. 获取 AppID 和 AppSecret

第二步：配置凭证

配置文件（多账号支持）

配置文件路径：~/.wechat-publish-pro/config.yaml

accounts:
  # 账号1（设为默认）
  default:
    name: 技术公众号
    app_id: wx89c409208d11dc5b
    app_secret: your_app_secret_here
  # 账号2
  health:
    name: 养生公众号
    app_id: cli_a95b243c6278dccf
    app_secret: your_app_secret_here
  # 可添加更多账号...
default_account: default

ai:
  provider: qwen
  api_key: your_api_key
  base_url: ""
  model: ""

环境变量（单账号兼容）

export WECHAT_APP_ID=your_wechat_app_id
export WECHAT_APP_SECRET=your_wechat_app_secret

# AI 去痕（可选）
export AI_API_KEY=your_api_key
export AI_PROVIDER=qwen  # openai, qwen, zhipu, doubao, minimax, moonshot, hunyuan, yi

第三步：设置 IP 白名单

把运行机器的 IP 添加到微信公众号后台白名单：

1. 登录 https://mp.weixin.qq.com/
2. 设置与开发 → 基本配置 → IP白名单

使用方式

通过 OpenClaw 直接发送

在 OpenClaw 对话中直接说：

• "把这篇文章发到公众号"
• "用蓝色主题发布"
• "帮我发布到微信，测试一下"
• "使用 AI 去痕发布"

命令行方式

Markdown 转换

# 基本转换（默认绿色主题）
wechat-publish-pro convert article.md

# 指定蓝色主题
wechat-publish-pro convert article.md --theme blue

# 指定输出文件
wechat-publish-pro convert article.md -o output.html

发布到微信

# 发布 HTML 文件到草稿箱（使用默认账号）
wechat-publish-pro publish article.html

# 指定账号发布
wechat-publish-pro publish article.html --account health

# 指定标题和封面
wechat-publish-pro publish article.html --title "文章标题" --cover cover.jpg

# 不使用 AI 去痕
wechat-publish-pro publish article.html --no-humanize

# 调整 AI 去痕强度
wechat-publish-pro publish article.html --intensity heavy

一站式：Markdown → 转换 → 发布

# 转换后直接发布
wechat-publish-pro convert article.md --theme blue -o /tmp/article.html
wechat-publish-pro publish /tmp/article.html --title "文章标题"

测试连接

wechat-publish-pro test

上传图片

wechat-publish-pro upload-image image.jpg

Markdown 格式

文件顶部可以包含 frontmatter：

---
title: 文章标题
cover: ./assets/cover.jpg
---

# 正文开始（frontmatter 有 title 时会自动移除）

你的内容...

可用主题

主题	风格
green	清新自然（绿色，默认）
blue	清新专业（蓝色）
purple	优雅神秘（紫色）
orange	温暖活力（橙色）
default	简洁清爽
simple	极简风格

AI 去痕

AI 去痕是可选功能，让文章读起来更自然。支持的 Provider：

Provider	说明
openai	OpenAI GPT 系列
qwen	通义千问（默认）
zhipu	智谱 GLM
doubao	豆包
minimax	MiniMax
moonshot	Moonshot
hunyuan	腾讯混元
yi	零一万物

去痕强度：

• light - 轻度：保持原文大部分内容，只做轻微调整
• medium - 中度：适度调整，保留核心内容
• heavy - 重度：大幅调整，使文章焕然一新

故障排查

错误：45166 (IP地址不在白名单中)

• 原因：运行机器的 IP 未添加到微信白名单
• 解决：登录公众号后台添加 IP 到白名单

发布成功但看不到文章？

• 原因：文章在草稿箱，需要审核发布
• 解决：草稿箱 → 选中文章 → 发布

架构说明

wechat-publish-pro/
├── src/wechat_publisher/
│   ├── cli.py         # CLI 入口
│   ├── config.py      # 配置管理
│   ├── converter/     # Markdown 转换模块
│   │   ├── wechat_style.py   # 微信公众号样式转换器
│   │   ├── themes.py          # 主题配置
│   │   └── theme_types.py     # 主题数据类型
│   ├── humanizer/     # AI 去痕模块
│   └── platforms/     # 平台适配器
│       └── wechat.py # 微信公众号 API
└── pyproject.toml

设计参考

本项目参考了以下优秀项目的设计思路：

• wenyan-cli: Markdown 转微信 HTML 的排版思路
• multi-writing-skills: 转换器和主题系统的实现

但本项目是完全独立实现的，不依赖任何外部代码。

作者

yuesf

许可证

MIT