# Set Up Libretto

Install Libretto and verify that it runs end to end.

## 1. Choose the target

Libretto is installed into the target Node.js package. Inspect the current directory and ask only for missing information:

- Create a new Libretto package? If so, where and what package name?
- Add Libretto to an existing Node.js package? If so, which package directory?

Use the target project's package manager conventions. If the user already chose a path, location, name, or package manager, treat that as binding.

## 2. Set up the package

For a new package:

```bash
npm create libretto@latest <package-name>
cd <package-name>
```

For an existing package:

```bash
npm install libretto
npx libretto setup
```

Use equivalent pnpm, yarn, or bun commands when appropriate.

## 3. Read installed guidance

After installation, make sure your current directory is the package directory. Then read the repo-level skill that setup copied into the project before creating or editing workflow code. Use whichever path exists:

```text
.agents/skills/libretto/SKILL.md
.claude/skills/libretto/SKILL.md
```

If neither path exists because the project has no `.agents/` or `.claude/` directory, create the appropriate agent directory and rerun `npx libretto setup`. The package also includes source docs under `node_modules/libretto/docs/`; use them only when the skill or task requires more detail.

## 4. Verify

For this smoke check, copy the workflow below directly; do not inspect the scaffolded example or read additional references unless validation fails.

Create a minimal smoke workflow at `src/workflows/scrape-page.ts` inside the package directory:

```ts
import { workflow } from "libretto";

export default workflow("scrape-page", async ({ page }) => {
  await page.goto("https://example.com");
  const title = await page.title();
  console.log(`Page title: ${title}`);

  return { title };
});
```

Run it:

```bash
npx libretto run src/workflows/scrape-page.ts --headless
```

## 5. Finish

After verifying Libretto is setup and working properly, summarize the steps taken and offer some sample browser automations you could build next, such as:

- Scrape recent posts from X/Twitter for a keyword, account, or list
- Monitor a product page for price or availability changes
- Fill out a repetitive web form from structured input data
- Collect search results from a directory, marketplace, or documentation site

## Important Instructions and Constraints to be Successful

- Fix only setup-related failures.
- Do not make unrelated changes or invent secrets.