notion-daily-review

name	notion-daily-review
description	Daily Notion knowledge review. Picks 2-3 knowledge points from user's Notion workspace, generates review summaries and deep-dive discussions. Trigger words: daily review, knowledge review, 每日回顾, 知识回顾, notion review, 今日知识

Notion Daily Knowledge Review

Intelligently selects 2-3 knowledge points from the user's Notion workspace, generates review summaries and deep-dive discussions for second-round learning and deeper understanding.

---

Prerequisites

This skill requires a Notion MCP Server to be connected. The recommended server is makenotion/notion-mcp-server.

Notion Integration Token Setup

1. Go to https://www.notion.so/my-integrations
2. Click "+ New integration"
3. Name it (e.g. "Cowork"), select your workspace, click Submit
4. Copy the Internal Integration Secret (starts with ntn_)
5. In each Notion page you want this skill to access: click "..." → "Connect to" → select your integration
6. Provide the token when the skill asks for it during initialization

Important: You must connect the integration to your top-level pages. Child pages inherit the connection automatically.

---

Configuration

All configuration is stored in the memory system at C:\Users\h\AppData\Roaming\Claude-3p\local-agent-mode-sessions\41d16b5b-9b59-4a80-9410-c824e988ec92\00000000-0000-4000-8000-000000000001\spaces\3cb39822-49ab-425e-bd2a-14960cb69d97\memory\ as a file named notion_daily_review_config.md.

Configuration fields:

notion_token: "<PENDING_INIT>"         # ntn_xxx integration token
review_database_id: "<PENDING_INIT>"   # created during init
review_database_title: "📝 Daily Knowledge Review"
knowledge_root_pages: []               # list of {id, name} objects
excluded_pages: []                     # page names to skip
style_preset: "<PENDING_INIT>"         # concise_deep | academic | casual | socratic
language: "<PENDING_INIT>"            # e.g. zh-CN, en
last_reviewed_pages: []                # track recent reviews to avoid repeats

---

First-Run Initialization

When to run: Execute if notion_token is <PENDING_INIT> or the config file does not exist.

Phase 0a: Check Notion MCP Connection

Check if Notion MCP tools are available by attempting a search:

mcp__notion-mcp-server__search(query="", page_size=1)

If this tool is not found, inform the user:

"This skill requires the Notion MCP Server. Let me help you set it up. Please install it from: https://github.com/makenotion/notion-mcp-server Then provide your Notion Integration Token."

If the tool exists but returns an auth error, ask for the token.

Phase 0b: Discover Knowledge Pages

Step 1: Get workspace pages:

mcp__notion-mcp-server__search(query="", page_size=100)

Step 2: For each candidate root page, inspect children:

mcp__notion-mcp-server__retrieve-block-children(block_id="<root_page_id>", page_size=100)

Step 3: Classify pages into:

• Knowledge pages: Notes, articles, book reviews, concepts, research
• Non-knowledge pages: Planners, trackers, to-do lists, calendars, dashboards

Step 4: Present classification to user and ask for confirmation.

Step 5: Record knowledge_root_pages and excluded_pages.

Phase 0c: Style Preferences

Present the following options:

"How would you like your daily reviews? Pick a style:

A. Concise & Deep (简练有深度) Core summary in 2-3 sentences + extension/new perspectives. No fluff.

B. Academic (学术探究) Structured analysis with frameworks, theories, cross-disciplinary connections.

C. Casual (轻松对话) Friendly, conversational. Like chatting with a smart friend.

D. Socratic (苏格拉底式) Heavy on thought-provoking questions. Maximum thinking stimulation.

Also, what language? (e.g. Chinese/中文, English)"

Phase 0d: Create Review Database

mcp__notion-mcp-server__create-database(
  parent={"type": "page_id", "page_id": "<parent_page_id>"},
  title=[{"type": "text", "text": {"content": "📝 Daily Knowledge Review"}}],
  properties={
    "Name": {"title": {}},
    "Date": {"date": {}},
    "Source Sections": {"multi_select": {"options": []}},
    "Trigger": {
      "select": {
        "options": [
          {"name": "Recent Edit", "color": "blue"},
          {"name": "Conversation", "color": "green"},
          {"name": "Random", "color": "gray"}
        ]
      }
    }
  }
)

Record the returned review_database_id.

Phase 0e: Save Configuration

Save the config to memory:

config = {
    "notion_token": "<token>",
    "review_database_id": "<id>",
    "review_database_title": "📝 Daily Knowledge Review",
    "knowledge_root_pages": [{"id": "...", "name": "..."}],
    "excluded_pages": ["..."],
    "style_preset": "concise_deep",
    "language": "zh-CN",
    "last_reviewed_pages": []
}
# Write to memory/notion_daily_review_config.md

Inform user:

"✅ Initialization complete!

• Monitoring X knowledge sections
• Review style: [style]
• Say '每日回顾' or 'daily review' to trigger, or set up a cron job."

---

Execution Flow

Prerequisite: Config must be initialized (not <PENDING_INIT>).

Phase 1: Collect Context Signals

1a. Get Recently Edited Pages

mcp__notion-mcp-server__search(query="", page_size=20)

Extract pages edited within the last 3 days (compare last_edited_time to today).

1b. Check Memory for Recent Topics

Read the memory index (MEMORY.md) to find recent conversation topics. Extract key words as context signals.

---

Phase 2: Select Knowledge Points

Dynamic Page Discovery

For each entry in knowledge_root_pages:

mcp__notion-mcp-server__retrieve-block-children(block_id="<root_page_id>", page_size=100)

Collect all child_page and child_database blocks. Filter out excluded pages and the review database.

Selection Strategy

Pick 2-3 knowledge points:

1. Context-related (1-2): Search using context signals:

mcp__notion-mcp-server__search(query="<keywords>", page_size=10)

Exclude non-knowledge pages, review database, and recently reviewed pages.

2. Random (1): Pick one from the knowledge list deterministically:
   • Use date +%j (day of year) modulo total knowledge sections count
   • This gives varied but reproducible daily picks

Read Content

For each selected page:

mcp__notion-mcp-server__retrieve-block-children(block_id="<page_id>", page_size=100)

For database-type sections, use:

mcp__notion-mcp-server__query-database(database_id="<db_id>", page_size=20)

Extract text from all block types.

---

Phase 3: Generate Review Content

Style Templates

concise_deep:

• Core summary: 2-3 sentences, no fluff
• Deep-dive: 150-250 words. New angles, cross-domain connections, 1-2 specific thought-provoking questions

academic:

• Core summary: Structured with key concepts
• Deep-dive: 200-300 words. Frameworks/theories, cross-disciplinary analysis, further reading

casual:

• Core summary: Conversational retelling
• Deep-dive: 150-250 words. Relatable examples, engaging tone

socratic:

• Core summary: Brief, 1-2 sentences
• Deep-dive: 200-300 words. Layered questions building to one challenging question

Output Format

## 🧠 Daily Knowledge Review — YYYY-MM-DD

Today's X knowledge points | [Why these were chosen]

---

### 📌 [Knowledge Point Title]
> Source: [Section] · Trigger: [Recent Edit / Conversation / Random]

**Core Review**
[Summary]

**Deep Dive**
[Extended discussion]

---

Writing Guidelines

• Use the configured language
• Include cross-references to other Notion pages when natural
• Questions should be specific and challenging
• Avoid verbatim repetition — add synthesis value
• Total output should not exceed 800 words

---

Phase 4: Output

4a. Write to Notion

mcp__notion-mcp-server__create-page(
  parent={"type": "database_id", "database_id": "<review_database_id>"},
  properties={
    "Name": {"title": [{"text": {"content": "YYYY-MM-DD Daily Knowledge Review"}}]},
    "Date": {"date": {"start": "YYYY-MM-DD"}},
    "Source Sections": {"multi_select": [{"name": "<section>"}, ...]},
    "Trigger": {"select": {"name": "<primary_trigger>"}}
  },
  children=[
    // heading_2, callout, heading_3, paragraph, divider blocks
  ]
)

Block structure per knowledge point:

[
  {"object":"block","type":"heading_2","heading_2":{"rich_text":[{"type":"text","text":{"content":"📌 Title"}}]}},
  {"object":"block","type":"callout","callout":{"rich_text":[{"type":"text","text":{"content":"Source: ... · Trigger: ..."}}],"icon":{"emoji":"📍"}}},
  {"object":"block","type":"heading_3","heading_3":{"rich_text":[{"type":"text","text":{"content":"Core Review"}}]}},
  {"object":"block","type":"paragraph","paragraph":{"rich_text":[{"type":"text","text":{"content":"..."}}]}},
  {"object":"block","type":"heading_3","heading_3":{"rich_text":[{"type":"text","text":{"content":"Deep Dive"}}]}},
  {"object":"block","type":"paragraph","paragraph":{"rich_text":[{"type":"text","text":{"content":"..."}}]}},
  {"object":"block","type":"divider","divider":{}}
]

4b. Update Review History

After successful write, update last_reviewed_pages in config (keep last 7 days, remove older entries).

4c. Send in Chat

Output the formatted review directly in the reply.

---

Important Notes

1. Exclude non-knowledge pages: Skip planners, trackers, to-do lists, daily logs, review database
2. Avoid repeats: Check last_reviewed_pages in config; skip pages reviewed in last 7 days
3. Handle empty content: If a page has < 50 characters, skip and pick another
4. Database-type sections: Use query-database instead of retrieve-block-children
5. Cross-references: Connect knowledge points to each other naturally
6. Error handling: If an API call fails, retry once after 2 seconds. If still failing, skip that step
7. Total length: Keep review output under 800 words across all knowledge points

---

Trigger Phrases

Any of these activate this skill:

• "每日回顾" / "知识回顾" / "今天回顾什么"
• "帮我回顾 Notion 知识"
• "daily review" / "notion review" / "knowledge review"
• Cron/scheduled task automatic trigger