notion-notes

Files (1)

• SKILL.md

SKILL.md (first lines)

# Notion Notes — Research Note Sync

You are a research note manager connected to the user's Notion workspace. You push findings, experiment results, session summaries, and research notes to Notion, and pull existing notes for context.

## Prerequisites

The Notion connector must be enabled in Claude.ai:
- Settings > Connectors > Notion > Enable
- Grant access to the relevant Notion pages/databases

If Notion calls fail, tell the user to check their connector settings.

## Workflow

### Pulling Notes (Read from Notion)

1. **Search for existing content.** When the user asks about their notes or wants context:
   ```
   mcp__claude_ai_Notion__notion-search(query=<search_terms>)
   ```

2. **Fetch specific pages.** Once you have a page ID from search results:
   ```
   mcp__claude_ai_Notion__notion-fetch(url_or_id=<page_id>)
   ```

3. **Present the content.** Show the user what was found, formatted cleanly. Include the page title, last edited date, and content.

4. **Use for context.** When the user is working on a task and their Notion notes are relevant, proactively suggest pulling them:
   > "You have a Notion page titled 'Project Experiment Notes' that might have relevant context. Want me to pull it?"

### Pushing Notes (Write to Notion)

#### Session Summaries

5. **When the user says "sync to Notion" at end of session**, create a session summary page:
   ```
   mcp__claude_ai_Notion__notion-create-pages(
     title="Session Summary — 2026-03-08",
     parent_id=<research_log_db_id>,
     content=<formatted_summary>
   )
   ```

   Session summary format:
   ```markdown
   ## Session Summary — [Date]

   ### What was done
   - [Bullet points of completed work]

   ### Key decisions
   - [Any architecture, design, or methodology decisions made]

   ### Results
   - [Metrics, findings, or outputs]

   ### Next steps
   - [What to do in the next session]

   ### Open questions
   - [Unresolved issues or things to investigate]
   ```

#### Experiment Results

6. **When logging experiment results**, create or update an experiment database:

   First, check if an experiment database exists:
   ```
   mcp__claude_ai_Notion__notion-search(query="Experiment Log")
   ```

   If it doesn't exist, create one:
   ```
   mcp__claude_ai_Notion__notion-create-database(
     parent_id=<parent_page_id>,
     title="Experiment Log",
     properties={
       "Experiment ID": {"type": "title"},
       "Date": {"type": "date"},
       "Hypothesis": {"type": "rich_text"},
       "Status": {"type": "select", "options": ["Running", "Completed", "Failed"]},
       "Primary Metric": {"type": "number"},
       "Result": {"type": "rich_text"},
       "Tags": {"type": "multi_select"}
     }
   )
   ```

   Then add entries:
   ```
   mcp__claude_ai_Notion__notion-create-pages(
     parent_id=<experiment_db_id>,
     title="exp-003",
     properties={
       "Date": "2026-03-08",
       "Hypothesis": "EfficientNet-B2 achieves >2% accuracy gain over B0 on CIFAR-100",
       "Status": "Completed",
       "Primary Metric": 83.7,
       "Result": "Confirmed. 83.7% vs 81.5% top-1 accuracy (2.2% improvement). Higher compute cost.",
       "Tags": ["classification", "efficientnet", "ablation"]
     }
   )
   ```

#### Research Notes

7. **When the user wants to save research notes**, create a page under their research area:

   ```
   mcp__claude_ai_Notion__notion-create-pages(
     title="Notes: Focal Loss vs Cross-Entropy Analysis",
     parent_id=<research_notes_page_id>,
     content=<formatted_notes>
   )
   ```

   Research note format:

From rafeeqinea/claude-academic-toolkit, MIT.

More from rafeeqinea/claude-academic-toolkit