What are element refs?

Element refs are unique identifiers generated during a snapshot that allow the agent to interact with buttons, inputs, and links using short, predictable aliases.

Can it handle websites that require login?

Yes, it supports multi-session management and the ability to save or load authentication states using session files.

Is it compatible with modern JavaScript frameworks?

Absolutely. It includes advanced 'wait' commands for network idle states and specific selectors to handle dynamic content in React, Vue, and other SPAs.

How does this skill reduce context usage by 93%?

It processes the page into an accessibility tree and uses short element references (like @e1) instead of the full HTML DOM, sending only essential interactive data to the AI.

Does it support capturing visual data?

Yes, the skill includes a screenshot tool that can save images to disk or return them as base64 strings for visual verification by the AI.

Browser Automation

Name: Browser Automation
Author: ruvnet

byruvnet

•

12,774

Browser Automation

Automates web navigation and interaction using AI-optimized accessibility snapshots to significantly reduce token usage.

About

The Browser Automation skill enables Claude agents to perform complex web-based tasks by utilizing a specialized agent-browser that translates web pages into lightweight accessibility trees. By replacing bulky DOM structures with compact element references, it reduces context window consumption by up to 93%, allowing agents to navigate, click, fill forms, and scrape data with high precision and lower costs. It is ideal for developers building autonomous workflows that require interacting with real-world websites, managing multiple browser sessions, and capturing visual feedback via screenshots.

Key Features

Interactive-only filtering to focus the AI on actionable UI elements
Multi-session support for parallel agent workflows and state persistence
AI-optimized accessibility snapshots with 93% context reduction
Simplified element reference system (@e1, @e2) for deterministic interactions
Advanced wait conditions for dynamic content and network idle states
12,774 GitHub stars

Use Cases

Automated end-to-end testing and visual verification of web applications
Automating multi-step authentication and dashboard navigation flows
Efficient web scraping and data extraction from complex SPAs

About

Key Features

Interactive-only filtering to focus the AI on actionable UI elements
Multi-session support for parallel agent workflows and state persistence
AI-optimized accessibility snapshots with 93% context reduction
Simplified element reference system (@e1, @e2) for deterministic interactions
Advanced wait conditions for dynamic content and network idle states
12,774 GitHub stars

Use Cases

Automated end-to-end testing and visual verification of web applications
Automating multi-step authentication and dashboard navigation flows
Efficient web scraping and data extraction from complex SPAs