Can I maintain authenticated sessions across commands?

Yes, the skill covers session management, cookie storage, and custom headers, allowing you to maintain logged-in states and bypass repetitive authentication flows.

Does it support real-time debugging of browser actions?

Absolutely. It supports a headed mode for visual debugging, CDP integration for developer tools access, and viewport streaming for real-time monitoring of agent actions.

What is the main advantage of using refs in agent-browser?

Refs (@e1, @e2) provide deterministic element selection by mapping accessibility tree elements to short IDs, which reduces token usage for LLMs and increases reliability compared to fragile CSS selectors.

How does agent mode benefit AI developers?

Agent mode provides machine-readable JSON output and optimized snapshots, making it significantly easier for AI models to parse web states and execute actions with high precision.

Agent Browser Automation

Name: Agent Browser Automation
Author: partme-ai

bypartme-ai

•

ブラウザ自動化

Automates web browser interactions using a CLI tool optimized for AI agents and deterministic element selection.

概要

This skill provides comprehensive guidance for using agent-browser, a high-performance CLI tool developed by Vercel Labs for autonomous web navigation. It enables AI agents to interact with web pages using deterministic 'refs', semantic locators, and accessibility trees, ensuring reliable automation that traditional CSS selectors often lack. By supporting features like JSON agent mode, persistent sessions, and Chrome DevTools Protocol (CDP) integration, it bridges the gap between LLM reasoning and real-world web interaction, making it an essential tool for developers building autonomous web agents or complex scraping workflows.

主な機能

Deterministic element selection using refs (@e1, @e2) and semantic locators
Web page snapshots with accessibility trees for enhanced LLM context
37 GitHub stars
Advanced debugging with headed mode, CDP integration, and live streaming
Agent mode with JSON output for seamless integration with AI workflows
Persistent session management and authenticated state handling via headers

ユースケース

Building autonomous AI agents that perform web research, data entry, or testing
Creating reliable web scrapers that leverage accessibility trees for robust data extraction
Automating complex form filling and multi-step web interactions via command line

概要

主な機能

Deterministic element selection using refs (@e1, @e2) and semantic locators
Web page snapshots with accessibility trees for enhanced LLM context
37 GitHub stars
Advanced debugging with headed mode, CDP integration, and live streaming
Agent mode with JSON output for seamless integration with AI workflows
Persistent session management and authenticated state handling via headers

ユースケース

Building autonomous AI agents that perform web research, data entry, or testing
Creating reliable web scrapers that leverage accessibility trees for robust data extraction
Automating complex form filling and multi-step web interactions via command line