browser-automation

name	browser-automation
description	This skill should be used when the user asks to "test in browser", "open a webpage", "take a screenshot", "fill a form", "E2E test", "verify in browser", "check the UI", "test the frontend", or needs to interact with a web application. Provides browser automation capabilities via the `agent-browser` CLI.

Browser Automation Skill

Use agent-browser CLI for browser automation, E2E testing, and web interactions.

Quick Start

# Open a webpage
agent-browser open https://example.com

# Take a screenshot
agent-browser screenshot /tmp/screenshot.png

# Get interactive elements
agent-browser snapshot -i

Common Workflows

1. Test a signup/login form

agent-browser open https://app.example.com/signup
agent-browser snapshot -i                        # See form elements
agent-browser fill @e1 "[e-mail]"        # Fill email
agent-browser fill @e2 "password123"             # Fill password
agent-browser click @e3                          # Click submit
agent-browser wait 2000                          # Wait for response
agent-browser screenshot /tmp/result.png         # Capture result
agent-browser get url                            # Check final URL

2. Verify page content

agent-browser open https://example.com
agent-browser snapshot                           # Get page content
agent-browser get title                          # Get page title
agent-browser get text @e1                       # Get element text

3. Debug network requests

agent-browser open https://example.com
# Fill and submit form...
agent-browser eval "JSON.stringify(performance.getEntriesByType('resource').filter(r => r.name.includes('api')).map(r => ({name: r.name})))"

Element Selection

• Refs: @e1, @e2 (from snapshot output)
• CSS: #id, .class, button[type=submit]
• Text: text=Login, text="Submit Form"

Key Commands

Command	Description
open <url>	Navigate to URL
snapshot -i	Get interactive elements
screenshot <path>	Capture screenshot
fill <selector> <text>	Fill input field
click <selector>	Click element
wait <ms>	Wait milliseconds
get url	Get current URL
get title	Get page title
eval <js>	Execute JavaScript

Tips

1. Always snapshot first before interacting with elements
2. Use -i flag to see only interactive elements
3. Take screenshots to verify state after actions
4. Element refs change after page updates - re-snapshot
5. Use sessions for multi-step workflows: agent-browser --session name open ...

When to Use

Use browser automation for:

• E2E testing signup/login flows
• Verifying UI changes
• Testing form submissions
• Checking page rendering
• Debugging frontend issues

Don't use when:

• Simple API test suffices (use curl)
• Checking static content (use WebFetch)