WeChat Publisher

Execute this workflow after article markdown already exists (for example from paper2wechat).

Inputs

Required:

• •Markdown file path, typically .paper2wechat/<paper_id>/outputs/<paper_id>.md

Optional:

• •WeChat AppID / AppSecret (for image upload and draft creation)
• •whether to upload images
• •whether to create draft
• •draft metadata: title/author/digest/thumb_media_id/content_source_url
• •theme: clean / card / tech / minimal / wechat-classic / ai-insight（推荐 AI 公众号）
• •keep first H1 in draft body: --keep-h1-in-draft (default removes first H1 to avoid duplicated title)

Credential sources (choose one):

1. •Environment variables (recommended for local runs):

export WECHAT_APP_ID="..."
export WECHAT_APP_SECRET="..."

1. •CLI args:

python .agents/skills/wechat-publisher/scripts/publish_wechat.py \
  --input-md ".paper2wechat/<paper_id>/outputs/<paper_id>.md" \
  --upload-images \
  --app-id "..." \
  --app-secret "..."

Do not commit secrets to git.

Default token cache path:

• •<workspace_root>/.wechat-token.json (shared across different papers)

Step 1: Resolve Markdown Assets

Run:

python .agents/skills/wechat-publisher/scripts/publish_wechat.py --input-md ".paper2wechat/<paper_id>/outputs/<paper_id>.md"

This baseline run does not require API credentials. It generates WeChat-friendly HTML and a copy helper page:

• •.paper2wechat/<paper_id>/outputs/<paper_id>.wechat.html
• •.paper2wechat/<paper_id>/outputs/<paper_id>.wechat.paste.html
• •.paper2wechat/<paper_id>/outputs/<paper_id>.wechat.themes.html (local theme switch preview: clean/card/tech/minimal/wechat-classic/ai-insight)

If default style is not preferred, switch theme:

python .agents/skills/wechat-publisher/scripts/publish_wechat.py \
  --input-md ".paper2wechat/<paper_id>/outputs/<paper_id>.md" \
  --theme card

Step 2: Upload Local Images To WeChat (Optional)

If markdown contains local image links and you want them available in WeChat backend:

WECHAT_APP_ID="..." WECHAT_APP_SECRET="..." \
python .agents/skills/wechat-publisher/scripts/publish_wechat.py \
  --input-md ".paper2wechat/<paper_id>/outputs/<paper_id>.md" \
  --upload-images

Expected artifacts:

• •image-map.json in the same outputs/ directory
• •image-hash-map.json in the same outputs/ directory (dedupe by file hash)
• •html files with image links rewritten to WeChat-hosted URLs

Image upload dedupe:

• •If image file path exists in map and content hash unchanged, uploader reuses existing URL and does not re-upload.
• •If image content changes, uploader uploads again and refreshes mapping.

Step 3: Create Draft In WeChat Backend (Optional)

To directly create a draft in Official Account backend:

WECHAT_APP_ID="..." WECHAT_APP_SECRET="..." \
python .agents/skills/wechat-publisher/scripts/publish_wechat.py \
  --input-md ".paper2wechat/<paper_id>/outputs/<paper_id>.md" \
  --upload-images \
  --create-draft \
  --auto-thumb

--auto-thumb will pick a suitable image from markdown (prefer framework/overview-like figures), upload it as permanent material, and use the returned thumb_media_id automatically.

Title behavior:

• •Publisher tries full title first.
• •If WeChat returns title-length error (45003), it retries with shorter title candidates.

Expected artifact:

• •publish-result.json in outputs/ containing media_id

Notes

• •Keep this skill independent from paper2wechat; do not modify existing parser or writing flow.
• •For stable rendering, prefer uploading images first, then generate html.
• •If API mode is unavailable, *.wechat.paste.html remains the fallback path: open it in browser and copy rich text into editor.

Resources

• •scripts/publish_wechat.py: deterministic publisher entrypoint
• •references/config-example.env: local env variable template