| name | wechat-mp-publisher |
| description | Publish Markdown articles to WeChat Official Account draft box using a local config file. Use when user wants to push blog posts to 微信公众号, convert Markdown to WeChat format, or create 公众号草稿. |
| metadata | {"version":"0.6.0"} |
WeChat MP Publisher
Publish Markdown articles to WeChat Official Account (微信公众号) draft box.
Usage
uv run <skill_dir>/scripts/publish.py -f <markdown-file> [-u <source-url>] [-c <cover-image>] [-t <title-override>] [-a <author>] [--config <config-file>]
The script:
- Gets access_token from WeChat API
- Pre-processes Markdown: footnotes → superscript references + "参考文献" section, bold fix
- Extracts and uploads images to WeChat CDN
- Converts Markdown to WeChat-compatible HTML with inline styles
- Uploads cover image as permanent material
- Creates draft via
draft/add API
Options
-f, --file — Markdown file path (required)-u, --url — Original article URL (set as "阅读原文" link)-c, --cover — Cover image path (defaults to first image in article)-t, --title — Override title (defaults to frontmatter title)-a, --author — Author name (defaults to "张昊辰")
Configuration
The script reads a local JSON config file. Default path:
/root/.openclaw/config/wechat-mp-publisher.json
Optional overrides: --config <path> or WECHAT_MP_CONFIG=<path> for one command. Do not store WeChat credentials in OpenClaw env.vars.
{
"app_id": "<wechat official account app id>",
"app_secret": "<wechat official account app secret>"
}
Markdown Features Supported
- Footnotes:
[^N] references become green superscript numbers; definitions at bottom become styled "参考文献" section with small gray text. External links in footnotes are converted to plain text (WeChat filters external URLs).
- Bold: Handles
**text**immediately-followed patterns that markdown-it can't parse.
- Images: Local images uploaded to WeChat CDN; external URLs kept as-is.
- Code blocks: Dark theme (One Dark), uses
<section> instead of <pre>/<code> to bypass WeChat's forced pre-wrap.
- Lists: Converted to
<p> + bullet/number characters (WeChat adds block-level wrappers inside <li>, breaking inline formatting).
- Tables: Fully styled with borders and alternating header.
Style Design
Optimized for both light and dark mode reading:
- No tinted backgrounds (they invert poorly in dark mode)
- Green accent (#07c160) for h2 underline, superscript refs, bold text (#2e7d32)
- Code blocks use dark background that stays dark in dark mode
- Images have no box-shadow (shadows look odd when inverted)
Output
Draft created in WeChat MP backend (mp.weixin.qq.com). User confirms and publishes manually.
Known Gotchas
- Bold regex cross-line bug (v0.2.0 fix): The bold pre-processor uses
[^*\n]+? (not [^*]+?) to prevent matching across paragraph boundaries. Without the \n exclusion, **end of paragraph pairs with beginning of next**, corrupting entire sections.
- Orphaned
--- separators: If the source Markdown has a horizontal rule between content and footnote definitions, the footnote extraction leaves an orphan ---. The script cleans trailing \n---\n before rendering.
- WeChat external URL filtering: All
[text](url) links in footnotes are stripped to plain text because WeChat silently drops external links from article content.
- Inline code layout break (v0.3.0 fix): Inline
<code> elements without explicit display:inline may be rendered as inline-block on WeChat mobile, causing surrounding text to break with large gaps. The style now includes display:inline;word-break:break-all;line-height:inherit;.
- Bold preprocessor breaks list items (v0.4.0 fix): The bold regex
**text**： added a space before CJK punctuation (**text** ：), causing WeChat to render extra visual gaps in list items. Fixed by excluding CJK punctuation (U+3000-303F, U+FF00-FFEF) from the lookahead.
- Bold cross-boundary matching (v0.4.1 fix): Bold regex
\*\*([^*\n]+?)\*\* matched across two bold boundaries when the closing ** of one bold and the opening ** of the next were on the same line, producing literal ** in output. Fixed with negative lookahead (?![。！？.!?\s]).
- Code block
white-space:pre-wrap (v0.5.0 fix): WeChat mobile (especially iOS) forcibly overrides white-space:pre to pre-wrap on <pre>/<code> elements, causing code to wrap instead of horizontal scroll. Fixed by replacing <pre><code> with <section> elements — WeChat preserves <section> styles as-is.
- List items line break before colon (v0.5.0 fix): WeChat wraps text after styled elements inside
<li> in block-level <section>, causing a line break between bold labels and their descriptions. Fixed by converting to tags with bullet/number characters.
Limitations
- Personal subscription accounts cannot auto-publish via API (requires manual confirmation)
- Images must be uploaded to WeChat CDN (external URLs filtered)
- Content size limit: 2MB, <20k characters
- Each push creates a new draft; old drafts must be manually deleted
- WeChat title limit: 64 characters (auto-truncated with …)
