agent-browser

Automate browser tasks via headless CLI with ref-based element targeting.

1|1|Updated Dec 28, 2025

Search Tags:#automation #testing #cli #playwright #browser #web-scraping #ref-based

Ready to run

npx skills add https://github.com/watzon/claude-code --skill agent-browser-watzon

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-browser
Source: https://github.com/watzon/claude-code/tree/main/skills/agent-browser
Command: npx skills add https://github.com/watzon/claude-code --skill agent-browser-watzon

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill enables automated browser tasks such as navigation, form submission, login flows, and data extraction using a CLI with ref-based element targeting to avoid brittle selectors.

Core Features & Use Cases

Open URLs and manage pages in headless mode
Take accessible snapshots and identify stable element refs (@e1, @e2)
Perform interactions (click, fill, type, scroll) and wait for navigation or events
Extract text, HTML, and data for verification and scraping
Support session isolation for parallel tasks

Quick Start

Open a page and begin interacting with the agent-browser CLI to establish a session.

Frequently Asked Questions about agent-browser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I automate web scraping without brittle CSS selectors?▼

Headless browser automation uses ref-based element targeting to identify stable elements via accessible snapshots, avoiding brittle selectors during web scraping and multi-step workflows.

How do I automate login flows and form submission in headless mode?▼

Automate login flows and form submission by opening URLs in headless mode, taking accessible snapshots, and performing interactions like click and fill using stable element refs.

What is the best way to run parallel browser automation tasks?▼

Parallel browser automation tasks are supported through session isolation, which manages separate browser instances via a daemon to prevent state collision across dynamic websites.

Do I need to install a separate CLI and daemon for browser automation?▼

Yes, this approach requires installing the agent-browser CLI and a daemon to manage browser instances, handle accessible snapshots, and execute interactions for end-to-end testing.

How does ref-based element targeting work for dynamic websites?▼

Ref-based element targeting works by taking accessible snapshots of dynamic websites and assigning stable identifiers like @e1 and @e2 to elements for reliable click and fill interactions.