Function getBestPracticesPrompt

packages/mcp/src/prompts.ts:367–423

()

}

function getBestPracticesPrompt(): GetPromptResult {
  return {
    description: 'Best practices for Deepnote notebooks',
    messages: [
      {
        role: 'user',
        content: {
          type: 'text',
          text: `# Deepnote Notebook Best Practices

## Structure
- Start with a title (text-cell-h1) and introduction (markdown)
- Organize into sections with headers (text-cell-h2)
- Add markdown explanations before complex code blocks
- End with conclusions or next steps

## Interactivity
- Convert hardcoded values to input blocks:
  - File paths → input-text
  - Thresholds/parameters → input-slider
  - Categories → input-select
  - Toggles → input-checkbox
- Use meaningful variable names and labels

## Code Quality
- Keep code blocks focused on one task
- Use df.head() or display() to show dataframes
- Add error handling for file loading
- Import all dependencies at the start

## Documentation
- Explain the "why" not just the "what"
- Document expected input formats
- Note any external dependencies
- Add callouts for important warnings

## Tools Workflow

**For new notebooks:**
1. ${TOOL_NAMES.create} → Creates project and initial notebook
2. ${TOOL_NAMES.addBlock} → Adds content and interactivity incrementally
3. ${TOOL_NAMES.run} → Validates the notebook end-to-end

**For existing notebooks:**
1. ${TOOL_NAMES.read} include=["lint","dag"] → Find issues
2. ${TOOL_NAMES.cat} + ${TOOL_NAMES.editBlock} → Inspect and fix blocks
3. ${TOOL_NAMES.run} + ${TOOL_NAMES.snapshotLoad} → Verify behavior

**For analysis:**
1. ${TOOL_NAMES.read} include=["structure","stats"] → See structure
2. ${TOOL_NAMES.read} include=["dag"] → View dependencies
3. ${TOOL_NAMES.snapshotLoad} → Inspect outputs from previous runs`,
        },
      },
    ],
  }
}