# Setup

> Install and configure the notebooklm-py CLI for querying NotebookLM notebooks. Handles installation, authentication, notebook discovery, and project config file setup.

- **Type:** Skill
- **Install:** `agentstack add skill-platdrag-notebooklm-claude-skill-setup`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [platdrag](https://agentstack.voostack.com/s/platdrag)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [platdrag](https://github.com/platdrag)
- **Source:** https://github.com/platdrag/notebookLM-claude-skill/tree/master/skills/setup

## Install

```sh
agentstack add skill-platdrag-notebooklm-claude-skill-setup
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# NotebookLM Setup

This skill guides the user through installing and configuring the `notebooklm-py` CLI tool so that `/notebooklm:query` can query their NotebookLM notebooks.

### Step 1: Check if notebooklm-py is Already Installed

```bash
notebooklm --version
```

- If the command succeeds, report the version and skip to Step 3.
- If the command is not found, proceed to Step 2.

### Step 2: Install notebooklm-py

Ask the user which installation method they prefer:

**Option A — Lightweight (recommended if you can export browser cookies manually):**
```bash
pip install notebooklm-py
```
This installs the CLI and all query functionality. Authentication must be handled by manually exporting cookies from an existing browser session (see Step 4, Option B).

**Option B — With browser-based login:**
```bash
pip install "notebooklm-py[browser]"
playwright install chromium
```
This adds Playwright, which enables the `notebooklm login` command. It opens a dedicated Chromium instance to capture Google session cookies — regular browsers don't expose cookies to external programs, so Playwright's own browser is needed for automated cookie extraction.

If `pip` fails, try `pip3` or suggest the user activate the appropriate Python virtual environment. `pipx install notebooklm-py` is also an option for global CLI tools.

Verify installation:
```bash
notebooklm --version
```

### Step 3: Check Authentication Status

```bash
notebooklm auth check --json
```

- If auth is valid, report the status and skip to Step 5.
- If auth is invalid or missing, proceed to Step 4.

### Step 4: Authenticate

Offer two authentication methods:

**Option A — Browser login (requires `[browser]` extra and Playwright):**

Tell the user:
> Please run this command in your terminal (not here — it opens a browser window):
> ```
> notebooklm login
> ```
> 1. A Chromium window will open to the NotebookLM page
> 2. Log in with your Google account
> 3. Once the page loads, switch back to your terminal and press ENTER
>
> Come back here when you're done.

**Option B — Manual cookie export (no Playwright needed):**

Tell the user:
> You can export cookies from a browser where you're already logged into NotebookLM:
>
> 1. Open https://notebooklm.google.com in your browser (make sure you're logged in)
> 2. Use a cookie export extension (e.g., "EditThisCookie", "Cookie-Editor", or "Get cookies.txt")
> 3. Export all cookies for the `.google.com` domain as JSON
> 4. Create the file `~/.notebooklm/storage_state.json` with this structure:
> ```json
> {
>   "cookies": [ ...exported cookies array... ],
>   "origins": []
> }
> ```
>
> The critical cookies are: `SID`, `HSID`, `SSID`, `APISID`, `SAPISID`, `__Secure-1PSID`, `__Secure-3PSID`

Wait for the user to confirm they've completed authentication.

Then verify:
```bash
notebooklm auth check --json
```

If still failing, suggest re-trying or switching to the other authentication method.

### Step 5: Discover Notebooks

List the user's notebooks:

```bash
notebooklm list --json
```

Present the results as a readable list:
```
Your NotebookLM notebooks:
1. "Notebook Title" (ID: abc123...)
2. "Another Notebook" (ID: def456...)
...
```

If no notebooks exist, suggest:
> You don't have any notebooks yet. You can create one:
> - **Web**: Go to https://notebooklm.google.com and click "New Notebook"
> - **CLI**: `notebooklm create "My Knowledge Base"` then add sources with `notebooklm source add`

### Step 6: Configure the Project Config File

Read the existing config at `{project-root}/.claude/notebooklm-config.json` if it exists.

Ask the user which notebook(s) they want to configure for this project. For each selected notebook, ask for a short name (key) to use with the `--notebook` flag.

Ask the user which notebook should be the default (used when no `--notebook` flag is specified with `/notebooklm:query`).

Write/update the config file at `{project-root}/.claude/notebooklm-config.json`:

```json
{
  "defaultNotebook": "",
  "notebooks": {
    "": {
      "id": "",
      "description": ""
    }
  },
  "settings": {
    "mode": "concise",
    "responseLength": "default",
    "persona": null
  }
}
```

**Settings reference:**
- `mode`: One of `default`, `learning-guide`, `concise`, `detailed`
- `responseLength`: One of `default`, `longer`, `shorter`
- `persona`: Custom instruction string (e.g., "Act as a senior architect") or `null` for default

### Step 7: Verify End-to-End

Run a quick test query against the configured default notebook:

```bash
notebooklm ask "Summarize the main topics covered in this notebook" -n  --json --new
```

If successful, show the user a summary of the response and confirm:
> Setup complete. You can now use `/notebooklm:query ` to query your notebook.

If it fails, diagnose based on the error (auth, notebook not found, no sources, etc.) and guide the user to fix it.

## Post-Setup

After setup is complete, remind the user:
- `/notebooklm:query ` — Query the default notebook
- `/notebooklm:query  --notebook ` — Query a specific notebook
- Auth sessions expire every few weeks — re-run `notebooklm login` or re-export cookies when needed
- To add more sources: `notebooklm source add "" -n `

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [platdrag](https://github.com/platdrag)
- **Source:** [platdrag/notebookLM-claude-skill](https://github.com/platdrag/notebookLM-claude-skill)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-platdrag-notebooklm-claude-skill-setup
- Seller: https://agentstack.voostack.com/s/platdrag
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.