chrome-automation

Launch and control Chrome with AT-SPI2 accessibility for browser automation. Use when asked to start chrome, open browser, launch chrome, begin browser automation, or control web pages.

⭐5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026⬇️3downloads

When & Why to Use This Skill

This Claude skill enables seamless browser automation by launching and controlling Google Chrome via AT-SPI2 accessibility features. It empowers AI agents to perform complex web interactions, including navigation, element clicking, and text input, effectively bridging the gap between LLMs and live web environments for automated task execution and data retrieval.

Use Cases

  • Web Testing and QA: Automate the process of navigating through web applications to verify UI elements and functional workflows programmatically.
  • Dynamic Data Extraction: Access and retrieve data from modern, JavaScript-rendered websites that require human-like interactions to reveal content.
  • Automated Workflow Integration: Execute repetitive web-based tasks such as filling out complex forms, managing CMS dashboards, or generating online reports.
  • Browser-Based Research: Enable agents to browse the live web, interact with search engines, and gather real-time information from specific URLs autonomously.
namechrome-automation
descriptionLaunch and control Chrome with AT-SPI2 accessibility for browser automation. Use when asked to start chrome, open browser, launch chrome, begin browser automation, or control web pages.

Chrome Automation

Launch Chrome with accessibility features enabled for programmatic control via AT-SPI2.

Quick Start

Check if Chrome is Running

~/.claude/skills/chrome-automation/scripts/launch.sh status

Or manually:

pgrep -f "chrome.*no-sandbox" && echo "Running" || echo "Not running"

Launch Chrome

~/.claude/skills/chrome-automation/scripts/launch.sh [URL]

The script automatically checks if Chrome is already running.

Or launch manually:

export DISPLAY=:1
export GTK_MODULES=gail:atk-bridge
export NO_AT_BRIDGE=0
export GNOME_ACCESSIBILITY=1

google-chrome \
    --no-sandbox \
    --disable-gpu \
    --start-maximized \
    --force-renderer-accessibility \
    --no-first-run \
    "https://google.com" &

Restart Chrome

~/.claude/skills/chrome-automation/scripts/launch.sh restart [URL]

Control Chrome (after launching)

List clickable elements:

chrome-a11y list

Click an element:

chrome-a11y click "Button Name"

Navigate to URL:

chrome-a11y navigate "https://example.com"

Type text:

chrome-a11y type "search query"

Send keyboard shortcut:

chrome-a11y key "ctrl+t"   # New tab
chrome-a11y key "ctrl+l"   # Focus address bar

Stop Chrome

pkill -9 chrome

Important Notes

  1. Must run in background with & to avoid blocking
  2. --force-renderer-accessibility is required for AT-SPI2 control
  3. --no-sandbox is required when running as root
  4. Wait 3-5 seconds after launch before using chrome-a11y

Prerequisites

Ensure these are installed:

  • TigerVNC running on :1
  • AT-SPI2 packages: gir1.2-atspi-2.0, at-spi2-core, python3-gi
  • The chrome-a11y tool in PATH

For detailed documentation, see REFERENCE.md.