Can I use this skill for websites that require login?

Yes, the skill supports session management, allowing agents to perform login flows and persist authentication states using session save/load capabilities.

Does it support modern single-page applications (SPAs)?

Yes, it includes robust 'wait' commands for network idle states, specific text, or element appearance, making it highly effective for dynamic SPAs built with React, Vue, or Angular.

How does this skill reduce token usage in Claude?

It uses an AI-optimized snapshot system that converts the complex DOM into a simplified accessibility tree with short element references (like @e1), reducing context size by approximately 93% compared to raw HTML.

What is the difference between snapshots and screenshots?

A snapshot provides a text-based accessibility tree that Claude can read and interact with using references, while a screenshot captures a visual base64 image of the current page for visual verification.

Web Browser Automation

Name: Web Browser Automation
Author: EarthmanWeb

byEarthmanWeb

•

브라우저 자동화

Automates web browser interactions using AI-optimized accessibility snapshots to minimize token usage and improve agent accuracy.

The Web Browser Automation skill enables Claude Flow agents to navigate, interact with, and extract data from websites using a specialized agent-browser interface. By leveraging AI-optimized snapshots that replace the full DOM with compact element references (@e1, @e2), this skill reduces context consumption by up to 93%, enabling faster and more reliable agent performance. It supports multi-session management, semantic locators, and automated wait states, making it an essential tool for building autonomous web-based workflows, scraping dynamic content, or performing end-to-end testing within a conversational AI framework.

주요 기능

012 GitHub stars

02Semantic locators and CSS selector support for robust element targeting

03AI-optimized snapshots reducing context by 93% using element refs

04Integrated wait states for network idle, URL patterns, and text appearance

05Full interaction suite including click, fill, type, hover, and scroll

06Isolated multi-session support for complex parallel agent swarms

사용 사례

01Building autonomous agents that can navigate and interact with web-based SaaS tools

02Automating multi-step login and form submission workflows across web platforms

03Extracting structured data from modern JavaScript-heavy web applications

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add earthmanweb/claude-flow-plugin browser

For use in Claude.ai and ChatGPT

Download Skill