# Notion To Notebooklm

> Extracts Notion pages and subpages recursively to a single concatenated Markdown file ready for NotebookLM.

- Skill: `alfons0329/notion-to-notebooklm` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add alfons0329/notion-to-notebooklm`
- Raw SKILL.md: https://api.skillmd.com/api/skills/alfons0329/notion-to-notebooklm/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: Alfons0329 (https://skillmd.com/u/alfons0329)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/alfons0329/notion-to-notebooklm

---


# Notion to NotebookLM Skill

**User-invoked.** This skill reads a Notion tree with a live token and writes files to disk — start it by typing the command; nothing here fires it on its own.

Extracts a Notion page and all of its subpages recursively and saves them as a single concatenated Markdown file. Supports natural language human overrides (e.g., exclusions) by translating them via LLM prior to executing the export.

## Arguments
- `notion-page-url` (required): The Notion page URL or ID to work with.
- `output-path` (required): The path to save the output downloaded raw data (can be a directory or a specific file name ending in `.md`).
- `--human-override` (optional): Human instructions for customization, e.g. "Do not export subpage virtual memory", "Exclude page X".
- `--prompt-output-dir` (optional): The directory path to save the generated custom NotebookLM podcast prompt file.

## Core Workflow

### Step 1: Argument Validation
- Ensure both `<notion-page-url>` and `<output-path>` are provided.
- If missing, immediately prompt the user for the missing argument(s).

### Step 2: Human Override Translation (LLM Phase)
- If a `--human-override` value is provided, analyze the text to identify any page titles, subpage names, or page IDs that the user wants to exclude or customize.
- Translate the natural language override into a structured, comma-separated list of exclusions.
  - *Example:* `Do not export subpage virtual memory` -> `virtual memory`
  - *Example:* `Exclude pages named OS Review and memory-hierarchy` -> `OS Review,memory-hierarchy`
  - *Example:* If no exclusions are specified or override is empty, the list is empty.

### Step 3: Script Execution
- Locate the Python helper script at [notion_to_markdown.py](file:///shared/scripts/notion_to_markdown.py).
- Build the command line argument list. If exclusions were extracted from the override in Step 2, append `--exclude "<comma-separated-list>"` to the command.
- Execute the script using command execution tools (`run_command`, `Bash`, or terminal execution) in the workspace directory:
  ```bash
  python3 shared/scripts/notion_to_markdown.py "<notion-page-url>" "<output-path>" [--exclude "<exclusions>"]
  ```
- Ensure the command environment has the `NOTION_TOKEN` environment variable populated.

### Step 4: Generate NotebookLM-Actionable Artifacts (LLM Phase)
- Read the generated raw Markdown file.
- Synthesize a **Podcast-Friendly Study Guide**:
  - Restructure the content into a conversational Q&A/FAQ format.
  - Remove noisy raw syntax (such as large code block diff headers) and focus on key concepts and vivid analogies.
  - Save this guide file in the same directory as the raw markdown, adding the suffix `-podcast-guide.md` to the filename.
- Synthesize a **Customized NotebookLM Podcast Prompt**:
  - Create a tailored prompt directing the NotebookLM hosts to focus on the key systems concepts, analogies, and technical trade-offs of the generated guide.
  - Save this prompt inside a text file named `<page-slug>-podcast-prompt.txt` inside the directory specified by `--prompt-output-dir`. If `--prompt-output-dir` is not provided, save it in the same directory as the study guide.

### Step 5: Error Handling & Output
- If the script execution fails (non-zero exit code), immediately surface the script's stderr output and errors to the user. Do not attempt to guess or modify the results.
- If the script execution succeeds, report the success to the user, displaying the paths of:
  1. The raw concatenated Markdown file.
  2. The podcast-friendly study guide Markdown file.
  3. The customized podcast prompt text file.
  - Provide a summary of the pages processed.