wechat-to-notion

name	wechat-to-notion
description	Save WeChat public account articles to a Notion database. Use when user sends a mp.weixin.qq.com link and wants to save/archive it to Notion. Fetches title, cover image, body content (paragraphs, headings, images, code blocks, lists) and writes them as Notion blocks.
metadata	{"openclaw":{"emoji":"📰","requires":{"env":["NOTION_API_KEY"]},"primaryEnv":"NOTION_API_KEY"}}

wechat-to-notion

Save a WeChat article to Notion in three steps: fetch → analyze → save.

Configuration Check (Do This First)

Check if the Notion API key is configured:

echo ${NOTION_API_KEY:0:8}...

If missing, tell the user:

You haven't configured a Notion API key yet:

1. Go to https://notion.so/my-integrations → + New integration → copy the key (starts with ntn_)
2. Open your Notion database → ... → Connect to → select your integration
3. Set the key in your OpenClaw config — do not paste it into chat:

   openclaw config set skills.entries.notion.apiKey "ntn_xxx"
   

   OpenClaw will inject it as NOTION_API_KEY automatically.

⚠️ Never ask the user to send the API key as a chat message — it will be exposed in conversation logs.

Setup (One-time)

Ask if the user has an existing Notion database. If yes, use it directly. If no, ask for a parent page URL and create one:

curl -s -X POST https://api.notion.com/v1/databases \
  -H "Authorization: Bearer $NOTION_API_KEY" \
  -H "Notion-Version: 2025-09-03" \
  -H "Content-Type: application/json" \
  -d '{
    "parent": {"type": "page_id", "page_id": "<parent_page_id>"},
    "title": [{"type": "text", "text": {"content": "WeChat Articles"}}],
    "properties": {
      "Title": {"title": {}},
      "URL": {"url": {}},
      "Read Time": {"date": {}},
      "Tags": {"multi_select": {}},
      "Notes": {"rich_text": {}}
    }
  }'

Match field names to the user's language (e.g. Chinese users get Chinese field names).

Workflow

Step 1: Fetch article

python3 {skillDir}/scripts/fetch_wechat.py <wechat_url> > /tmp/wx_article.json

Step 2: Analyze (inline — reason directly, no subprocess)

Use the read tool to load /tmp/wx_article.json. Read the title and text content from blocks, then produce two outputs by reasoning directly:

Keywords (3–5):

• Lead with proper nouns: product names, tool names, frameworks, technical terms (e.g. Claude Code, MCP, OpenClaw, RAG)
• Add domain-specific concepts only when they add meaning (e.g. multi-agent orchestration, agentic workflow)
• Omit generic filler (e.g. "AI", "efficiency", "productivity", "tools")
• Comma-separated, preserve original casing

Comment (1 sentence, written in the user's language): Write a single sentence structured in two parts:

1. A tight, neutral distillation of the article's core argument — precise enough to serve as a standalone summary quote
2. A parenthetical editorial aside: a sharp, specific take on the piece — its quality, angle, or the author's intent. Be direct. No hedging.

• Example: "MCP bridges OpenClaw and Claude Code, eliminating manual context-passing between agents (well-executed tutorial — the headline actually holds up)"
• Keep it under 35 words

Step 3: Save to Notion

python3 {skillDir}/scripts/save_to_notion.py \
  /tmp/wx_article.json \
  <notion_db_url> \
  <wechat_url> \
  <read_time_iso8601+08:00> \
  "<kw1>,<kw2>,<kw3>" \
  "<comment>"

• read_time: current time in the user's local timezone as ISO 8601 with offset, e.g. 2026-03-12T14:00:00+08:00
• keywords: comma-separated string
• comment: the single-sentence comment from Step 2

The script auto-detects field names from the database schema by type (title, url, date, multi_select), writes all content blocks in batches of 100, and posts the comment to the Notion Comments panel.

Notes

• Cover image: extracted from og:image meta tag, inserted as the first block
• Rich text (bold/italic), code blocks, and lists are preserved by fetch_wechat.py
• Read time defaults to current system time with local UTC offset if omitted
• The comment appears in the Notion Comments panel, not in the page body
• Field names are language-agnostic — the script maps by type, not by name