An agent must inspect a live flow
Load the CLI's core workflow before navigating, snapshotting controls, or testing a form.
agent-browser Skill by Vercel Labs
Use the agent-browser Skill by Vercel Labs to connect an agent with browser workflows for the agent-browser CLI. Prepare the tool separately, load instructions that match its installed version, and choose the guidance for website interaction, exploratory testing, or supported Electron tasks.
Install and get started
Choose your agent to see how to get started. In Floatboat, copy a request that asks the agent to read the original instructions. Other options show the Skill installation command or the publisher’s native setup steps.
Choose an agent
Start in Floatboat
Copy the request into a Floatboat conversation
Read the original agent-browser Skill instructions at https://github.com/vercel-labs/agent-browser/blob/main/skills/agent-browser/SKILL.md. Explain its purpose, required tools, and installation options, then help me use it.Let the agent read the original instructions and check how to use them in your environment.
Download Floatboat ↗Try this request
Use agent-browser to test the sign-in flow. Load the current core instructions first, then report the exact step that fails.
About agent-browser Skill
The agent-browser Skill is a companion instruction published by Vercel Labs for the agent-browser browser-automation CLI. Its short entry file directs the agent to load core or specialized workflows from the installed CLI, keeping detailed instructions aligned with the tool version. Core guidance supports website interaction, while specialized entries address tasks such as exploratory testing and Electron applications. The Skill and CLI have separate roles: installing the instruction does not install the tool or its browser dependencies. Prepare those through the upstream setup guide, then load the matching workflow before acting on a page. Review the current page state and verify the result of each task.
Use Cases for the agent-browser Skill
Explore browser tasks where an agent needs the CLI’s matching operating instructions. Inspect website interactions, reproduce an issue through exploratory testing, or load specialized Electron guidance, with the required tool installed and the task scoped to the pages or application you intend to use.
A bug needs exploratory reproduction
The stub points to a specialist dogfood workflow for exploratory testing and quality review.
The target is an Electron app
Use the CLI's dedicated Electron guidance instead of assuming a normal web-page session.
How to Use the agent-browser Skill
Add the companion Skill to your agent and prepare the agent-browser CLI with its browser dependencies. Load the CLI’s core instructions or the relevant specialized workflow, inspect the current target, and verify that the resulting page or application state matches the requested task.
- 01
Add the companion Skill
Use the Skills installation options above to add the instruction to your agent. Keep this separate from setting up the browser tool.
- 02
Prepare the CLI and browser
Follow the upstream agent-browser setup for the CLI and its browser dependencies. Confirm the tool is available in the agent’s environment.
- 03
Load a workflow and verify
Read agent-browser skills get core or the matching specialist workflow. Inspect the current target, carry out the task, and verify the resulting state.
Explore More Agent Skills
Read another Skill guide and check its original source before using it.
agent-browser Skill FAQ
Does installing the Skill install the browser CLI?
No. The Skill adds a short instruction stub. The agent-browser CLI and browser setup are separate installations.
Where are the full browser instructions?
The installed CLI serves them through
agent-browser skills get core and specialist workflows such as dogfood or electron.Why does the repository look like a tool?
Because it is primarily a CLI tool. The SKILL.md is a companion component for AI agents; the two should not be classified as the same deliverable.
Can the Skill run a browser by itself?
No. The CLI performs navigation, snapshots, and interactions. The Skill tells the agent which CLI-served workflow to read first.
Why can a browser instruction stop matching the page?
Navigation and dynamic updates can replace controls or change page state. Inspect the current page again and retarget the action rather than repeating a stale element reference.
Can an agent submit any form once the browser is connected?
Browser access does not establish permission for every action. Confirm the user’s intended destination and action before sending messages, publishing material, purchasing, or changing account access.
How do I prepare the CLI separately from the Skill?
Follow the repository’s tool setup instructions. Its entry file lists npm i -g agent-browser followed by agent-browser install for the tool and browser setup. Then load agent-browser skills get core before browser tasks; check upstream requirements for your platform.