Drive a real Chrome via the DevTools Protocol — navigate, evaluate JS, screenshot, pick DOM elements interactively, and inspect cookies — for frontend testing and automation needing a visible browser.
---
name: browser-tools
description: Interactive browser automation via the Chrome DevTools Protocol. Use to interact with web pages, test frontends, or when a visible browser and human interaction are required.
---
CDP-based browser automation through bundled Node scripts, connecting to Chrome on port 9222.
## How to use
1. Run `npm install` once, then launch Chrome with remote debugging via `browser-start.js` (add `--profile` to reuse the user's cookies and logins).
2. Navigate with `browser-nav.js <url>` (`--new` for a new tab) and run code in the page with `browser-eval.js '<js>'` to extract data or inspect state.
3. Capture the viewport with `browser-screenshot.js` to verify UI; read cookies with `browser-cookies.js`.
4. When selectors are ambiguous, run `browser-pick.js "<instruction>"` to let the user click elements interactively and return their CSS selectors.
Full skill & source: https://github.com/badlogic/pi-skills/tree/90bb51cae36515a648515b633a81c0c6efc8c74d/browser-toolsSign in to rate and review this skill.
No reviews yet. Be the first to review this skill.