WeChat Publish Draft

Upload a formatted WeChat article HTML file and cover image to the WeChat Official Account draft box. This skill creates a backend draft only; it never clicks publish, free-publishes, or mass-sends without a separate explicit request and a separate safety review.

Inputs

Expected inputs:

• WeChat-ready HTML from wechat-layout.
• Article title from wechat-rewrite.
• Cover image from wechat-cover-image, or an existing WeChat cover media_id.
• Optional digest, author, source URL, account suffix, and comment settings.

If HTML, title, or cover/media_id is missing, ask for the missing item.

Existing Configuration

Reuse existing config from:

/Users/ziye/Library/Mobile Documents/com~apple~CloudDocs/wiki/30-39 Automation and Agents/31 Agent Workflows/31.11 WeChat Skill/skills/wechat-publisher/.env
/Users/ziye/Library/Mobile Documents/com~apple~CloudDocs/wiki/30-39 Automation and Agents/31 Agent Workflows/31.11 WeChat Skill/scripts/.env

The script also supports a local skill .env file and environment variables.

Required variables:

WECHAT_APPID=...
WECHAT_APPSECRET=...

For multiple accounts:

WECHAT_APPID_A=...
WECHAT_APPSECRET_A=...
WECHAT_APPID_B=...
WECHAT_APPSECRET_B=...

Do not print AppSecret, access_token, or full media IDs unless the user explicitly needs them. It is okay to report that credentials exist.

Workflow

Step 1: Validate Local Files

Check:

• HTML file exists.
• Cover image exists, unless --cover-media-id is provided.
• HTML is inline-style and suitable for WeChat editor import.
• Any local <img src="..."> paths are reachable from the HTML file directory or absolute paths.

Step 2: Get Access Token

Use the configured AppID/AppSecret to call:

GET https://api.weixin.qq.com/cgi-bin/token

If the request returns IP whitelist errors such as 40164, tell the user to add the current machine/server outbound IP to WeChat Official Account backend 设置与开发 -> 基本配置 -> IP白名单.

Step 3: Upload Inline Images

For local body images, call:

POST https://api.weixin.qq.com/cgi-bin/media/uploadimg

Replace local img src values with returned WeChat image URLs. Do not upload remote HTTP images; warn if remote non-WeChat image URLs remain.

Step 4: Upload Cover

If --cover-media-id is absent, upload the cover with:

POST https://api.weixin.qq.com/cgi-bin/material/add_material?type=image

Use the returned media_id as thumb_media_id.

Step 5: Create Draft

Call:

POST https://api.weixin.qq.com/cgi-bin/draft/add

Payload shape:

{
  "articles": [
    {
      "title": "...",
      "author": "...",
      "digest": "...",
      "content": "...",
      "content_source_url": "...",
      "thumb_media_id": "...",
      "need_open_comment": 1,
      "only_fans_can_comment": 0
    }
  ]
}

Use json.dumps(..., ensure_ascii=False).encode("utf-8") to avoid Chinese garbling.

Command

python3 /Users/ziye/.agents/skills/wechat-publish-draft/scripts/upload_draft.py \
  --html "<wechat-layout.html>" \
  --title "<article title>" \
  --cover "<cover-900x383.jpg>"

Optional:

  --account A \
  --author "AI比我快" \
  --digest "文章摘要" \
  --source-url "https://..." \
  --cover-media-id "<existing media_id>" \
  --no-open-comment \
  --dry-run

Safety Rules

• Create drafts only.
• Do not call freepublish/submit, mass send APIs, or UI publish buttons.
• Do not expose secrets in final output.
• Do not overwrite or delete existing drafts.
• If the user asks to publish publicly, stop and create a separate plan/checklist first.
• If WeChat returns an error, report the errcode, a short explanation, and the next action.

Output Shape

Return:

## 草稿箱结果
- 状态：成功/失败
- 草稿 media_id：<redacted or full if needed>
- 账号：默认/A/B

## 上传内容
- HTML：
- 封面：
- 正文图片数量：

## 后台检查
- 去微信公众号后台草稿箱检查标题、封面、正文排版。
- 只确认草稿，不自动发布。