Skip to content
Get packs

Guide · Agent skill

How to use the Browser Automation via Hive Browser CLI agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

This guide explains the hive.browser-automation agent skill, which teaches an AI agent to operate a real Chrome browser that you already have open, using the hive-browser command-line tool through a browser extension. It is written for readers who have never installed an agent skill before. By the end you will know what the skill changes, how to install it, how to ask for browser work in plain language, and what safety limits to keep in mind.

Browser Automation via Hive Browser CLIaden-hive/hive · Apache-2.0

Quick start

  1. 1Download and unzip. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
  2. 2Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill instead.
  3. 3Confirm the runtime and extension. Make sure the Hive runtime with hive-browser is installed and Chrome has the GCU Beeline extension connected.
  4. 4Run setup. Run hive-browser setup --json to confirm a browser is connected and to see any install steps it reports.
  5. 5Make your request. Describe the site, the task and the account in plain language, naming the profile label if you have several.
  6. 6Check the profile in the output. Confirm the JSON echoes the profile you intended before the agent continues.

What an agent skill is and what this one does

An agent skill is a folder with a SKILL.md file that gives an AI agent instructions for a specific kind of task. This skill covers browser automation. It tells the agent to drive the browser from the terminal by running hive-browser commands with --json through the terminal_exec tool, rather than through MCP tools.

The agent can take screenshots, click and type at proportional positions on the page, and read page elements. Coordinates are fractions of the viewport (0 to 1 on each axis), not pixels, so a target about 35% from the left and 20% from the top is passed as --coordinate 0.35,0.20. The source says this approach reaches inputs in Shadow DOM that selectors cannot see.

  • Connects to the Chrome you already run, via the GCU Beeline extension and Chrome DevTools Protocol (CDP)
  • Uses screenshot plus coordinate clicks when selectors fail
  • Includes rules for rich-text editors and CSP problems
  • Points to related LinkedIn skills for LinkedIn work

Install the skill (ZIP or Claude app upload)

Download the ZIP of the skill from this page and unzip it. For a personal setup, place the unzipped folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

  • Personal: ~/.claude/skills/
  • Project: .claude/skills/ inside your project folder
  • Claude apps: Settings → Capabilities → Skills

Check requirements before you start

The skill only works where its prerequisites exist. The source lists three: the Hive runtime with the hive-browser CLI, Chrome with the GCU Beeline extension connected via Chrome DevTools Protocol, and a terminal execution tool (terminal_exec).

If no browser is connected, run hive-browser setup --json. It reports the connection state and gives the user install steps. Do not try to launch Chrome yourself.

  • Hive runtime with hive-browser CLI
  • Chrome with the GCU Beeline extension connected via CDP
  • terminal_exec tool available to the agent
hive-browser setup --json

Ask for browser work in natural language

You do not need to type commands yourself. Describe the task and name the site and account. The agent loads this skill before any hive-browser command, because the skill is required first.

If you have several Chrome profiles connected, name the one to use. The agent passes it with --browser-profile <label> and checks that the JSON response reports the profile it actually used. Labels are shown in each profile's Hive extension side panel.

  • Example request: "Open the vendor portal in my work profile and take a screenshot of the invoices page."
  • Example request: "In profile acct-jpn, find the search box on this page and type my query."
  • Name the account whenever it matters, since omitting the profile uses the first-connected one
hive-browser status --json

Safety notes: scripts, network and logged-in sessions

This skill runs scripts and uses the network. The browser holds the user's logged-in sessions, so the agent acts inside those accounts and could post or message as you. Use it with supervision on sensitive accounts.

The skill forbids launching browser processes or killing Chrome, the bridge, or gcu processes. A timeout does not mean the browser is stuck. The agent is told to wait about 30 seconds and retry once with a smaller request, then close its own tabs, then report the failure.

  • Do not use it for headless or isolated scraping; it only attaches to your own Chrome
  • Review what the agent is about to post or send on logged-in accounts
  • Heavy hive-browser evaluate scripts can cause timeouts; the skill advises chunking them

What people use it for

  • Filling in a form on a logged-in web app that uses shadow DOM inputs selectors cannot reach
  • Taking screenshots of a page in a specific Chrome profile to check its current state
  • Running multi-agent work where several agents share one browser connection under shared lifecycle rules
  • Working on LinkedIn flows by loading the related hive.linkedin skills after this one

Frequently asked questions

How do I install the browser automation agent skill?

Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for personal use or a project's .claude/skills/ folder for one project. In the Claude apps, you can upload it under Settings → Capabilities → Skills.

What do I need to run this skill?

The source lists the Hive runtime with the hive-browser CLI, Chrome with the GCU Beeline extension connected via Chrome DevTools Protocol, and a terminal execution tool (terminal_exec). Without these, the instructions cannot be followed.

Is this skill free and what is its licence?

The source lists the licence for the aden-hive/hive repository as Apache-2.0. The source does not state pricing, so check the repository for any cost details.

Can the agent use a headless or separate browser instead?

No. The skill attaches only to your own running Chrome and forbids launching browser processes or using headless flags. If you need an isolated browser for scraping, this skill is not the right fit.

Why did my browser command time out?

According to the skill, a timeout is a transport issue, not a crashed browser. Often another agent's heavy hive-browser evaluate call is slowing the shared connection. Wait about 30 seconds and retry once with a smaller request.

What is the difference between this and LinkedIn skills?

This skill covers general browser control. For any LinkedIn flow, the skill says to load hive.linkedin-core first, then a capability skill such as hive.linkedin-messaging or hive.linkedin-connect.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers and automation users who need an agent to act inside a logged-in Chrome session, especially on pages with shadow DOM. It is a poor fit if you want a headless or isolated browser, or if you cannot supervise actions on sensitive accounts. You also need the Hive runtime and hive-browser CLI installed, or the instructions will not work.

Download the skill