wechat-article-fetcher

name	wechat-article-fetcher
description	Fetch WeChat public account articles from mp.weixin.qq.com into clean Markdown with local image assets. Use when the user asks to grab, archive, crawl, save, export, or convert a WeChat/微信公众号 article, especially when they want images downloaded locally, Markdown image links rewritten to an assets folder, or an offline copy of a WeChat article.

WeChat Article Fetcher

Overview

Use this skill to turn a WeChat public account article URL into a local Markdown file plus a sibling assets directory containing downloaded article images.

Prefer the bundled script instead of hand-copying image URLs. It uses the existing fetch-skill backend order for WeChat content, then adds deterministic local asset handling.

This skill does not own a repository. The output path decides where the archive is written. Resolve bundled scripts from the active skill directory, and do not use the spreadconfig skill source directory as an implicit archive destination.

Quick Start

Resolve the bundled script relative to this SKILL.md directory. Do not hardcode ~/.codex, /nix/store, or any other absolute installation path.

SKILL_DIR="<directory containing this SKILL.md>"
SCRIPT="$SKILL_DIR/scripts/fetch_wechat_article.py"

Run with an existing Python 3 interpreter when available:

python3 "$SCRIPT" "https://mp.weixin.qq.com/s/..." -o article.md

The script only uses the Python standard library. On NixOS, do not assume python3 is globally available. If no Python interpreter is on PATH, run through Nix:

nix shell nixpkgs#python3 --command python3 "$SCRIPT" "https://mp.weixin.qq.com/s/..." -o article.md

The default output layout is:

article.md
article_assets/
  image01.png
  image02.jpg

Workflow

1. Choose an output filename. For article IDs, prefer wechat_<id>.md.
2. Resolve scripts/fetch_wechat_article.py from the active skill directory, not from the current working directory or a hardcoded install path.
3. Run the script with the URL and output path using python3, or the Nix command above when Python is not on PATH.
4. If the run fails due to network sandboxing or Nix daemon access, rerun the same command with escalation approval.
5. Inspect the top and tail of the Markdown for WeChat UI noise.
6. Verify that no remote image CDN URLs remain:

rg -n "mmbiz\.qpic\.cn|!\[[^]]*\]\(https?://" article.md
find article_assets -type f | sort

Script Behavior

The script:

• unwraps common WeChat captcha redirect URLs via target_url;
• fetches content through WECHAT_API_URL or --wechat-api when configured, then Jina Reader, then defuddle.md, then raw HTML;
• removes common trailing WeChat UI text such as slide prompts and mini-program controls;
• adds an H1 title from frontmatter when the backend produced metadata but no visible title;
• downloads Markdown and HTML image URLs into the assets directory;
• rewrites Markdown image references to relative local paths;
• fails on image download errors by default so offline archives do not silently keep broken remote links.

Use --best-effort-images only when a partial archive is acceptable. Use --no-images only when the user explicitly does not need local images.

Options

Common options:

python3 "$SCRIPT" URL \
  -o output.md \
  --assets-dir output_assets \
  --wechat-api http://localhost:3000 \
  --timeout 30 \
  --best-effort-images

Environment:

• WECHAT_API_URL: optional wechat-article-exporter base URL, for example http://localhost:3000.

Quality Bar

Before finishing, report:

• Markdown file path.
• Assets directory path.
• Number of downloaded images.
• Whether remote image links remain.

If a backend returns noisy Markdown, clean only obvious WeChat page chrome. Do not paraphrase or rewrite article content unless the user asks for a summary or rewrite.