Notion Inbox AI Reading Notes Generator

Generate deep reading notes for articles under a Notion Inbox page using the Notion REST API directly. No MCP plugin required. Default fast mode: batch parallel processing, no web search.

Language

Match the user's language: respond in the same language the user uses.

Configuration

Credentials (.env)

Check .env existence (priority order):

# Project-level
test -f .nby-skills/nby-notion-reading-notes/.env && echo "project"
# User-level
test -f "$HOME/.nby-skills/nby-notion-reading-notes/.env" && echo "user"

Required keys:

NOTION_TOKEN=your_notion_api_token

Preferences (EXTEND.md)

Check EXTEND.md existence (same priority order):

test -f .nby-skills/nby-notion-reading-notes/EXTEND.md && echo "project"
test -f "$HOME/.nby-skills/nby-notion-reading-notes/EXTEND.md" && echo "user"

Supported keys:

Example EXTEND.md:

inbox_id: your_inbox_page_id
target_id: your_target_page_id
batch_size: 3
note_language: zh

First-Time Setup

When neither .env nor EXTEND.md is found, guide the user:

Notion Reading Notes — First-Time Setup

Step 1: API Token
  → Visit https://www.notion.so/my-integrations
  → Create an integration, copy the token
  → Share target Notion pages with the integration

Step 2: Page IDs
  → Open Inbox page in browser, copy the 32-char ID from URL
  → Open target page (e.g., "AI精读") in browser, copy its ID

Where to save?
  A) Project-level: .nby-skills/nby-notion-reading-notes/
  B) User-level: ~/.nby-skills/nby-notion-reading-notes/

After collecting values, create both .env (token only) and EXTEND.md (page IDs and preferences).

Value Priority

CLI arguments > EXTEND.md (project) > EXTEND.md (user) > Skill defaults

API Quick Reference

Important: Use POST /move for moving pages. PATCH to set parent is silently ignored by Notion.

Core Workflow (Batch Parallel)

Round 1: List all Inbox child pages, record IDs and titles
── Per batch (configurable, default 3) ──
Round 2: Read all articles in batch in parallel (N concurrent Bash calls)
Round 2.5 (if needed): Fetch external link content via Jina WebFetch
Round 3: Analyze batch → write JSON + append notes in parallel (N Bash calls)
Round 4: Move to target page in parallel (N Bash calls)
── Next batch ──

Key principles:

• No detection needed: Everything in Inbox is unprocessed
• No web search: Extract resource links only from article content
• Batch size: Balance speed with Notion API rate limit (3 req/s)
• Append then move: Append notes (Round 3) before moving (Round 4) to avoid 429 errors
• Non-blocking failures: Log failures, continue to next article

Execution Steps

Round 1: List Inbox Children

curl -s "https://api.notion.com/v1/blocks/${INBOX_ID}/children?page_size=100" \
  -H "Authorization: Bearer ${NOTION_TOKEN}" \
  -H "Notion-Version: 2022-06-28" \
  | python3 -c "
import sys,json; d=json.load(sys.stdin)
for r in d['results']:
    if r['type']=='child_page': print(f\"{r['id']} → {r['child_page']['title']}\")
"

Split results into batches per configured batch_size.

Round 2: Read Article Content in Parallel

For each article in the batch, launch a concurrent Bash call:

curl -s "https://api.notion.com/v1/blocks/{PAGE_ID}/children?page_size=100" \
  -H "Authorization: Bearer ${NOTION_TOKEN}" \
  -H "Notion-Version: 2022-06-28" \
  | python3 -c "
import sys,json; d=json.load(sys.stdin)
types=['paragraph','heading_1','heading_2','heading_3','bulleted_list_item','numbered_list_item','quote','callout','toggle','code']
for r in d['results']:
    t=r['type']
    if t in types:
        print(''.join(x.get('plain_text','') for x in r[t].get('rich_text',[])))
    elif t=='bookmark': print(f\"[link: {r[t].get('url','')}]\")
    elif t=='image': print('[image]')
    elif t=='table': print('[table]')
if d.get('has_more'): print(f'has_more, cursor={d[\"next_cursor\"]}')
"

• Pagination: Use &startcursor={nextcursor} for additional pages
• Link-only pages: Mark as "⚠️ Insufficient content, skipped"

Round 2.5 (If Needed): Fetch External Content

For articles whose body is primarily a bookmark link, fetch via Jina:

WebFetch: https://r.jina.ai/{LINK_URL}

Launch multiple WebFetch calls in parallel. Mark as "insufficient content" on failure. Skip for articles with adequate content.

Round 3: Analyze + Append Notes in Parallel

Analyze all articles in batch, generate note JSON. Launch concurrent Bash calls:

cat <<'NOTEJSON' > /tmp/notion_note_{PAGE_ID}.json
{note JSON content}
NOTEJSON
curl -s -X PATCH "https://api.notion.com/v1/blocks/{PAGE_ID}/children" \
  -H "Authorization: Bearer ${NOTION_TOKEN}" \
  -H "Notion-Version: 2022-06-28" \
  -H "Content-Type: application/json" \
  -d @/tmp/notion_note_{PAGE_ID}.json

For JSON template details, refer to references/note-template.md.

Round 4: Move to Target Page in Parallel

curl -s -X POST "https://api.notion.com/v1/pages/{PAGE_ID}/move" \
  -H "Authorization: Bearer ${NOTION_TOKEN}" \
  -H "Notion-Version: 2022-06-28" \
  -H "Content-Type: application/json" \
  -d "{\"parent\":{\"type\":\"page_id\",\"page_id\":\"${TARGET_ID}\"}}"

Completion Report

✅ AI Reading Notes Complete: X succeeded, Y failed
1. 《Title A》- Core: xxx
2. 《Title B》- Core: xxx

Note Generation Guidelines

• Design table columns dynamically based on article type (technical/opinion/tools/methodology)
• Bold first column with "annotations": {"bold": true}
• 🔗 Resources: Conditional — only when article contains tool/library links; extract only, no web search
• 💡 Insights: Required — 2-4 actionable bullet points, personalized, concise

For full JSON templates: references/note-template.md

Error Handling

• Insufficient content: Move to target, append "⚠️ Insufficient content, skipped"
• Pagination: Continue with startcursor until hasmore=false
• Single failure: Log, skip to next
• API 429: Wait 2s, retry up to 3 times
• File conflicts: Each article uses /tmp/notionnote{PAGE_ID}.json

Additional Resources

• references/note-template.md — Notion Block JSON templates, formatting rules, optional sections