web-automation-runner

Run browser automation tasks on any website using Playwright. Scraping, form filling, data extraction, screenshots — all through natural language instructions.

Usage

Scrape [what] from [URL]
Extract [data description] from [URL]
Fill and submit [form details] on [URL]
Take a screenshot of [URL]
Monitor [element/text] on [URL] and alert when it changes

What It Does

The agent uses a real Chromium browser to:

• Navigate to any URL
• Click buttons, fill forms, select dropdowns
• Wait for elements or text to appear
• Extract structured data from pages
• Take screenshots (viewport or full page)
• Handle dialogs (alerts, confirmations, prompts)
• Run custom JavaScript on the page

Examples

• "Scrape all job listings from remoteok.com - need title, company, salary, and apply link"
• "Fill out the contact form on example.com/contact using: name=John, email=john@mail.com, message=Hi"
• "Take a screenshot of this dashboard and save it to dashboard.png"
• "Extract all product names and prices from amazon.com/s?k=laptop"
• "Login to [site] with username X and password Y, then scrape my order history"

Data Formats

Extracted data is returned as structured text or written to files:

• CSV for tabular data
• JSON for structured records
• TXT for plain text

Limitations

• Cannot bypass CAPTCHAs or login barriers (except credentials you provide)
• Rate-limited sites may need delays
• Heavy pages may take longer to load
• No PDF generation from dynamic content

Notes

• Uses Playwright with Chromium (headless)
• All actions are logged with timestamps
• Screenshots saved to workspace with descriptive names