微信文章格式转换

微信公众号文章抓取与转换，支持 Markdown/HTML/Text/JSON/Excel 五种格式。 触发条件：用户发送微信文章链接（https://mp.weixin.qq.com/s/xxx）、 多链接（换行分隔）、或「帮我转这篇文章」「保存微信文章」等表达。 格式选择前必须询问用户，支持单选或多选。

Security Scan

The skill largely implements what it advertises (fetch and convert WeChat articles) but there are coherence issues (undeclared dependencies, no install instructions) and a network-safety concern (it will fetch arbitrary asset URLs found in articles, which can reach internal hosts).

Purpose & Capability

The name/description match the included scripts: fetch_article.py and batch_fetch.py implement fetching mp.weixin.qq.com articles and saving Markdown/HTML/Text/JSON/Excel. The code saves output to the user's Desktop as described. However, SKILL.md lists only html5lib as a dependency while the code imports bs4 (BeautifulSoup) and likely uses additional libraries for Excel output — a missing dependency declaration is an inconsistency.

Instruction Scope

Instructions and scripts confine activity to fetching articles and saving converted files, and they require asking the user for output format first as stated. But the implementation will download every asset URL found in article HTML (CSS, images, background URLs) with no domain restriction; that means the tool can make requests to arbitrary hosts (including internal network addresses) if such URLs are embedded in the article.

Install Mechanism

There is no install spec (instruction-only), but code files are included. The SKILL.md declares html5lib only; the code imports bs4 (beautifulsoup4) and probably an Excel library (not declared). The lack of explicit dependency/install instructions is a coherence/operational gap: users may have to install packages manually and may unknowingly install from unvetted sources.

Credentials

The skill requests no environment variables, no credentials, and no config paths. File writes are limited to user-controlled output directories (~/Desktop/...), which aligns with the stated purpose.

Persistence & Privilege

The skill is not always-enabled and does not request elevated persistence. Autonomous invocation is allowed (platform default) but there are no additional privilege escalations in the package.

What to consider before installing

What to consider before installing: - Functional coherence: The scripts do what the description says (download article HTML and assets, convert and save). However SKILL.md lists only html5lib while the code imports BeautifulSoup (bs4) and likely an Excel library — confirm and install required Python packages (beautifulsoup4, html5lib, and the Excel writer such as openpyxl/xlsxwriter) from trusted sources before running. - Network safety: The tool downloads all asset URLs it finds in the article HTML without domain restriction. A malicious or crafted article could include links to internal network addresses (127.0.0.1, 192.168.x.x) or other unexpected hosts; consider running the tool in a network-restricted or sandboxed environment if you are concerned about internal service access. - Filesystem impact: It writes output under your home Desktop by default. If you allow it, it will create directories and save binary assets — review the save location and run in a directory you control. - Code review/validation: Review the rest of fetch_article.py (the truncated portion) and save_article implementation to confirm there is no telemetry, hidden remote endpoints, or unexpected behavior (e.g., uploading files elsewhere). The provided excerpts show no evidence of exfiltration, but the file is truncated so complete review is recommended. - Safe testing: Run first on a single, benign article in an isolated environment; inspect downloaded assets and logs to confirm behavior before bulk use. If you want, I can: (1) list precise Python packages this code requires by scanning the rest of fetch_article.py, (2) extract and show the save_article function and Excel code for review, or (3) suggest command-line steps to create a safe virtualenv and run one test fetch.

将微信文章转换为离线的 Markdown/HTML/Text/JSON/Excel，无需登录，直接抓取，图片和样式本地化保留。

微信公众号文章抓取

触发条件

用户发送：

• 微信文章链接：https://mp.weixin.qq.com/s/xxx
• 多个链接（换行或空格分隔）
• 「帮我转这篇文章」「保存这篇微信文章」「下载这篇文章」

每次导出前必须询问格式（必须）

用户: https://mp.weixin.qq.com/s/xxx
→ 询问: "请选择导出格式（支持单选或多选，例如：1 或 1,3,5 或 markdown,html）"

可选格式：

序号	格式	扩展名	说明
1	Markdown	.md	带标题/作者/日期/来源链接
2	Excel	.xlsx	表格，含元信息+正文（行高300px）
3	HTML	.html	独立网页，CSS+图片本地化，浏览器直接打开
4	纯文本	.txt	无格式纯文字
5	JSON	.json	结构化，含完整元数据

格式选择解析（不区分大小写）：

• 单选：2 / excel / Excel
• 多选：1,3,5 / markdown,html,json / 1 和 3

HTML 格式详解

HTML 是离线阅读效果最好的格式，结构如下：

文章标题_文章ID/
├── index.html        # 完整网页（CSS引用+图片引用）
└── assets/           # 资源文件夹
    ├── *.css         # 样式文件（18个，共约3MB）
    └── *.jpg/png/webp # 图片文件

生成策略：

• 保留微信原始 HTML 结构（<!DOCTYPE><html><head><body>）
• 下载所有 CSS/图片到 assets/，用时间戳文件名
• 替换所有资源 URL 为本地相对路径 href="./assets/xxx"
• CSS/图片按内容 hash 去重（相同资源只存一份）
• #js_content 的 visibility:hidden 样式自动移除（防空白页）
• #js_article_bottom_bar 底部栏自动保留
• MINIMAL_CSS 内联到 <style> 保底样式（字体/行高/图片自适应）
• 空 img 标签（视频/音频占位符）自动清理
• 无需登录，直接请求 mp.weixin.qq.com，Referer 必须带域名

CSS 处理策略：

• CSS 总和 ≤ 500KB → 内联到 HTML <style>（单文件方便分享）
• CSS 总和 > 500KB → 保存到 assets/ 用 <link> 引用（避免 HTML 臃肿）

已知限制：

• 阅读量/点赞/评论数：微信 API 需要登录态 cookies，纯抓取无法获取
• 极少数被删除或违规文章无法抓取（返回错误）

单篇处理流程

1. 接收用户链接
2. 必须先询问格式
3. 调用脚本（每个选定格式一次）
4. 保存到 ~/Desktop/文章标题_文章ID/
5. 汇报文件名和大小

批量处理流程

1. 接收多链接（换行分隔）
2. 必须先询问格式
3. 每篇每格式各抓一次，保存到 ~/Desktop/微信文章批量/
4. 汇报成功数/失败数

核心脚本

# 单篇（单格式）
python skills/wechat-article/scripts/fetch_article.py <url> [format] [output_dir]
# format: markdown | html | text | json | excel

# 批量（多链接）
python skills/wechat-article/scripts/batch_fetch.py <urls_file> [formats_csv]

实现细节

• 依赖：html5lib（纯 Python，无 C 扩展，pip install html5lib）
• 无需登录：直接请求 mp.weixin.qq.com，带 Chrome UA + Referer
• 图片处理：优先用 data-src（懒加载），同步到 src；空 img 标签自动移除
• 文件名：文章标题前80字符 + 文章ID，避免重复
• Excel：深蓝表头，交替底色，含边框
• 并发下载：CSS/图片并行（8线程），超时12秒/资源