browserbase/skills-old

Overview

This skill is a complete guide for creating, testing, and deploying browser automation functions using the browse CLI and Browserbase Functions. It walks through interactive site exploration, initializing a function project, a required package.json fix, writing Playwright-based automation, local testing, and production deployment. Follow the steps to avoid the common build error and get reliable scheduled or webhook-driven automations.

How this skill works

You use the browse CLI to inspect pages interactively (open, snapshot, screenshot, click, fill) and then scaffold a Function project with the SDK. The function connects to a Browserbase browser session via Playwright (connectOverCDP), runs automated flows in index.ts, and returns JSON-serializable results. Local dev runs at a local HTTP endpoint for quick iteration; publishing builds and uploads the function to Browserbase for production invocation via API keys or webhooks.

When to use it

• Automating repetitive website workflows (form fills, navigation, downloads).
• Scraping structured data from pages or multi-page sequences.
• Building Browserbase Functions to run on a schedule or via webhook.
• Testing and debugging site interactions interactively before coding.
• Deploying headless browser tasks that require Playwright access.

Best practices

• Immediately fix package.json after init: add description, main, packageManager, pinned versions, and devDependencies to avoid build failures.
• Explore the site interactively with browse open / snapshot / screenshot to find robust selectors before coding.
• Return only JSON-serializable data and include timestamps and counts for observability.
• Test locally using pnpm bb dev index.ts and invoke with curl to verify behavior and logs.
• Use specific selectors (data attributes preferred), add logging, validate params, and handle errors with try/catch.

Example use cases

• Hacker News top-10 scraper that returns title, url, points, author, and comments.
• Authenticated workflows: sign in, navigate to dashboard, and extract user data.
• Parameterized scraper that accepts url and selector to collect lists of text content.
• Multi-page title collector that visits an array of URLs and returns titles.
• Scheduled data collection deployed to Browserbase and invoked via webhook or cron.