Skip to content
Get packs

Browser Automation via DOMShell

Lets an agent browse and click through web pages by navigating Chrome's accessibility tree like files.

Source Download skill
Repository
HKUDS/CLI-Anything
Stars
58,084
Licence
Apache-2.0
Files
1 file

What it does

Once installed, the agent can open web pages in your Chrome browser and move through the page's structure using filesystem-style commands such as ls, cd, cat, grep, click and type. Instead of looking at screenshots, it reads the page's accessibility tree, a text map of buttons, inputs, headings and landmarks. It can fill in forms and click buttons on your behalf. It relies on a companion Chrome extension and a Node.js server called DOMShell, plus a Python command-line tool.

Great for

  • Developers who want an AI agent to test or automate a web app they are building
  • Analysts who need to pull structured content from pages that have no API
  • People building agent workflows that need repeatable, scriptable browser steps

Skip it if

  • Logging into sensitive accounts (banking, email) through an agent: the browser shares cookies and login state across sessions, so the agent could act with your real credentials
  • Browsing untrusted sites while the agent acts autonomously: page text can contain instructions meant to hijack the agent, as the skill itself warns
  • Users who cannot install Node.js, Python 3.10+ and a Chrome extension, since setup is required

Try asking

  • “Open example.com and list the main navigation links”
  • “Find the Login button on this page and click it”
  • “Fill in the email field with my address and submit the form”
  • “Search the current page for the word pricing and show me the surrounding elements”
Step-by-step guide: how to use Browser Automation via DOMShellInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.

    skills.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip skills.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skills.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Start with page open and fs ls / to see the page structure before clicking anything
  • Use --json output when the agent parses results, and keep a human reviewing actions on untrusted sites
  • Set CLI_ANYTHING_BROWSER_BLOCK_PRIVATE=true if you do not want the agent reaching localhost or your home network, and use a separate Chrome profile for sensitive work

What's inside

  • SKILL.md

Redistributed unmodified from HKUDS/CLI-Anything under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.