Can I handle authenticated sessions securely?

Yes, the skill includes an Auth Vault to store credentials using encryption and state management features to save and load cookies or localStorage across different sessions.

How does Agent Browser interact with elements?

It uses a snapshot system that assigns short, stable references like @e1 or @e2 to interactive elements, allowing AI agents to click, fill, or select elements without needing complex CSS selectors.

What security measures are available for AI agents?

It provides domain allowlists to restrict navigation, action policies to gate destructive commands, and content boundaries to help the LLM distinguish tool output from untrusted page content.

Does it support mobile browser testing?

Yes, Agent Browser can launch and interact with Safari on iOS Simulators (on macOS) or physical devices, supporting mobile-specific gestures like swipes and taps.

How does visual diffing work?

The 'diff' command allows you to compare current page states against previous snapshots or baseline images to verify that an action had the intended effect or to detect UI regressions.

Agent Browser Automation

Name: Agent Browser Automation
Author: decodetool

bydecodetool

0•

ブラウザ自動化

Automates browser workflows including web navigation, form interaction, data extraction, and visual testing for AI-driven tasks.

Agent Browser is a robust CLI-based tool designed to bridge the gap between AI agents and the web, providing a specialized environment for navigating websites and interacting with DOM elements via unique reference IDs. It facilitates complex browser operations like managing authenticated sessions through a secure Auth Vault, persistent state handling, and parallel session support. Whether you are automating routine web tasks, scraping data, or conducting end-to-end testing, this skill offers advanced features like visual diffing, mobile Safari simulation via iOS Simulators, and granular security controls such as domain allowlists and action policies to ensure safe and reliable automation.

主な機能

01Mobile Safari simulation with iOS Simulator and physical device support

02Integrated Auth Vault and state management for secure, persistent sessions

030 GitHub stars

04Advanced visual and accessibility-tree diffing for step-by-step verification

05Granular security controls including domain allowlists and action policies

06Interactive DOM interaction using unique element reference IDs (@e1, @e2)

ユースケース

01Automating multi-step web workflows such as form filling and account management

02End-to-end web application testing and visual regression monitoring

03Automated web scraping and structured data extraction from complex sites

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add decodetool/agent-browser agent-browser

For use in Claude.ai and ChatGPT

主な機能

01Mobile Safari simulation with iOS Simulator and physical device support

02Integrated Auth Vault and state management for secure, persistent sessions

030 GitHub stars

04Advanced visual and accessibility-tree diffing for step-by-step verification

05Granular security controls including domain allowlists and action policies

06Interactive DOM interaction using unique element reference IDs (@e1, @e2)

ユースケース

01Automating multi-step web workflows such as form filling and account management

02End-to-end web application testing and visual regression monitoring

03Automated web scraping and structured data extraction from complex sites