wechat-public-article-archive

WeChat Public Article Archive

Archive only user-provided public article links or local HTML. Keep the access result truthful: anonymous GET is an attempt, not a guarantee of access.

Enforce the boundary

• Accept only https://mp.weixin.qq.com/... article URLs and local HTML the user saved.
• Use no login, browser profile, cookie, token, QR scan, proxy, endpoint adapter, JavaScript execution, or account-history discovery.
• Do not follow redirects. Treat redirects, 401, 403, 429, CAPTCHA, and login/verification pages as needs_manual_browser.
• Download only HTTPS images from mmbiz.qpic.cn, mmbiz.qlogo.cn, or wx.qlogo.cn; record any rejected or failed image in the manifest.
• Cap downloads at 20 images; record excess images as skipped / asset_limit_reached and report asset_partial.
• Keep article text, images, and real URLs out of Git by default. Archiving is not permission to republish content.

Archive a known URL

From this skill directory, run:

node scripts/wechat-public-archive.mjs ingest \
  --source "SOURCE" \
  --url "https://mp.weixin.qq.com/..." \
  --archive-dir "/absolute/local/archive"

Read the JSON receipt. Report archived or asset_partial as completed local archives; report needs_manual_browser, network_error, or invalid_input exactly as returned.

Import user-saved HTML

When the receipt is needs_manual_browser, let the user save the page HTML in their own browser, then import it locally without opening a browser or reading credentials:

node scripts/wechat-public-archive.mjs ingest \
  --source "SOURCE" \
  --html "/absolute/path/article.html" \
  --captured-url "https://mp.weixin.qq.com/..." \
  --archive-dir "/absolute/local/archive"

--captured-url is optional for offline imports. Without it, data-URI images can still be archived; remote images are retained as failed manifest entries because their base URL is unknown.

Inspect the archive

Each successful archive contains:

ARCHIVE_ID/
├── original.html       # source bytes
├── content.html        # extracted article body markup
├── content.md          # extracted title and plain text
├── metadata.json       # source, URL, time, and SHA-256
├── assets.json         # per-image provenance and outcome
└── assets/             # only allowlisted image bytes