Allows to interact with web pages by performing actions such as clicking buttons, filling out forms, and navigating links. It works by remote controlling Google Chrome or Chromium browsers using the Chrome DevTools Protocol (CDP). When Claude needs to browse the web, it can use this skill to do so.
Inherits all available tools
Additional assets for this skill
This skill inherits all available tools. When active, it can use any tool Claude has access to.
tools/eval.jstools/nav.jstools/package.jsontools/pick.jstools/screenshot.jstools/start.jsMinimal CDP tools for collaborative site exploration.
./tools/start.js # Fresh profile
./tools/start.js --profile # Copy your profile (cookies, logins)
Start Chrome on :9222 with remote debugging.
./tools/nav.js https://example.com
./tools/nav.js https://example.com --new
Navigate current tab or open new tab.
./tools/eval.js 'document.title'
./tools/eval.js 'document.querySelectorAll("a").length'
./tools/eval.js 'JSON.stringify(Array.from(document.querySelectorAll("a")).map(a => ({ text: a.textContent.trim(), href: a.href })).filter(link => !link.href.startsWith("https://")))'
Execute JavaScript in active tab (async context). Be careful with string escaping, best to use single quotes.
./tools/screenshot.js
Screenshot current viewport, returns temp file path
./tools/pick.js "Click the submit button"
Interactive element picker. Click to select, Cmd/Ctrl+Click for multi-select, Enter to finish.