Web to WeChat

Web to WeChat

抓取任意网页链接的内容，AI 智能整理排版，自动生成封面图，发布到微信公众号草稿箱。

User-Facing Promise

Accept requests like:

• "帮我抓取这个链接的内容，发到微信公众号"
• "把这篇文章转载到我的公众号"
• "抓取 https://xxx.com/article 发布到微信"
• "把这个网页内容整理成公众号文章"
• "帮我收藏这篇文章到我的公众号"

Return a published draft in the WeChat draft box, not a proposal.

First-Time Setup

1. Install Python Dependencies

python -m pip install requests beautifulsoup4 html2text markdown Pillow

2. Install Companion Skills

This skill depends on two companion skills from ClawHub:

clawhub install anything-to-wechat
clawhub install file-to-wechat

• anything-to-wechat — provides publish_to_wechat.py (WeChat API publishing)
• file-to-wechat — provides md_to_wechat_html.py (Markdown → WeChat HTML)

3. Configure WeChat Credentials

You need a WeChat Official Account (服务号 or 订阅号 with API access).

Get your credentials:

1. Log in to https://mp.weixin.qq.com/
2. Go to: 设置与开发 → 基本配置
3. Copy your AppID and reset/view AppSecret
4. Add your server's public IP to the IP白名单

Set environment variables (recommended):

On macOS / Linux:

export WECHAT_APP_ID="your_appid_here"
export WECHAT_APP_SECRET="your_appsecret_here"

On Windows (PowerShell):

[Environment]::SetEnvironmentVariable("WECHAT_APP_ID", "your_appid_here", "User")
[Environment]::SetEnvironmentVariable("WECHAT_APP_SECRET", "your_appsecret_here", "User")

No environment variables? The publish script will prompt you interactively on first run.

Prerequisites

Workflow

Phase 1: Collect Input

If the user has NOT provided a URL, ask using AskUserQuestion:

Question: "请提供你想抓取并发布到微信公众号的网页链接"
Options:
  - "粘贴 URL"
  - "搜索关键词后选择文章"

If the user already provided a URL, skip and proceed.

Phase 2: Scrape Web Content

IMPORTANT: Always use UTF-8 encoding on Windows.

Primary: Use scrape_web.py script (structured extraction)

python "/scripts/scrape_web.py" \
  --url "" \
  --output "/raw_article.md" \
  --json

The --json flag outputs structured data including title, author, date, cover URL, and Markdown content.

Fallback: Use WebFetch tool (for JS-rendered pages)

If scrape_web.py fails (e.g., the page requires JavaScript rendering), use the built-in WebFetch tool:

WebFetch(url="", prompt="Extract the full article content including: title, author, publish date, and the complete article text. Return in a structured format.")

Then manually compose the Markdown from the WebFetch output.

After scraping: Read the output. Inspect the content structure and quality. Proceed to Phase 3.

Phase 3: AI Content Reformatting

This is the key quality step. The agent should:

1. Read the scraped Markdown content
2. Rewrite/reformat the content for WeChat publishing:
   • Ensure clean heading hierarchy (H1 → H2 → H3)
   • Fix broken formatting (tables, lists, code blocks)
   • Remove navigation artifacts, ads, social media buttons
   • Ensure paragraphs flow naturally
   • Add section breaks where appropriate
   • Keep the original meaning and facts intact
3. Output a clean, well-structured Markdown file

Style options (ask user or auto-detect):

Save the reformatted Markdown:

Write the reformatted content to /article.md using the Write tool.

Phase 4: Generate Cover Image

Use the ImageGen tool with a prompt derived from the article's topic and content.

• Save as wechat_cover.png in the workspace.
• Size: 1024x768 (WeChat cover ratio 4:3).
• Make it visually compelling and relevant to the article topic.

Phase 5: Compress Cover Image

WeChat requires cover images (thumb_media_id) to be under 64KB.

python "/scripts/compress_image.py" \
  --input "/wechat_cover.png" \
  --output "/wechat_cover_compressed.jpg" \
  --max-size 64

If the original cover is already under 64KB (rare for PNG), this step can be skipped. But always run it to be safe — it won't enlarge files.

Fallback: If compress_image.py fails to reach 64KB, use ImageGen to regenerate a simpler cover image (fewer details, simpler composition) and try again.

Phase 6: Convert to WeChat HTML

Use md_to_wechat_html.py from the file-to-wechat skill:

python "/scripts/md_to_wechat_html.py" \
  --input "/article.md" \
  --output "/wechat_article.html" \
  --title ""

This generates WeChat-compatible inline-style HTML with Clockless design tokens.

Phase 7: Publish to WeChat Draft Box

On Windows, use Python subprocess to pass environment variables:

python -c "
import os, subprocess, sys
os.environ['WECHAT_APP_ID'] = ''
os.environ['WECHAT_APP_SECRET'] = ''
result = subprocess.run([
    sys.executable,
    r'/scripts/publish_to_wechat.py',
    '--file', r'/wechat_article.html',
    '--title', '',
    '--cover', r'/wechat_cover_compressed.jpg',
    '--digest', '',
    '--source-url', ''
], capture_output=True, text=True, encoding='utf-8')
print(result.stdout)
print(result.stderr)
"

Or with environment variables already set:

python "/scripts/publish_to_wechat.py" \
  --file "/wechat_article.html" \
  --title "" \
  --cover "/wechat_cover_compressed.jpg" \
  --digest "" \
  --source-url ""

Credentials: The script reads from WECHAT_APP_ID / WECHAT_APP_SECRET env vars. If not set, it prompts interactively.

Phase 8: Confirm & Handoff

Report success with Media ID and link to https://mp.weixin.qq.com/.

Tell the user: "文章已发送到你的微信公众号草稿箱，请登录微信公众平台审核后一键发布。"

Include:

• Original source URL
• Article title
• Content summary
• Media ID

Site Compatibility

Error Handling

Copyright & Attribution

IMPORTANT: When republishing web content:

1. Always include the original source URL in the article metadata
2. Preserve the original author's name when available
3. Add a disclaimer line at the end: "本文内容来源于[原作者名称]，原文发布于[来源网站]。如有侵权请联系删除。"
4. Use --source-url when publishing to add a "Read More" link
5. Consider using "精华摘要" or "深度改写" style instead of exact copy to avoid copyright issues

Script Reference

Configuration