vibium-browser-automation_skill

Overview

This skill documents how to use the Vibium browser automation tool via the MCP protocol inside Cursor IDE. It provides concrete examples for launching a browser, navigating pages, finding and interacting with elements, taking screenshots, and closing the session. The guide highlights limitations and practical workarounds to integrate Vibium with other browser tools. It focuses on reliable patterns for web scraping, testing, and form automation within Cursor.

How this skill works

Vibium exposes simple operations: browser_launch, browser_navigate, browser_find, browser_click, browser_type, browser_screenshot, and browser_quit. Each operation waits for appropriate page state (load, clickable, editable) and returns element metadata (tag, text, bounding box) where relevant. Screenshots are saved by default to the Pictures/Vibium directory. Use Vibium for in-IDE programmatic browser control and combine it with other tools when features are missing.

When to use it

• Automating navigation and interactions inside Cursor IDE
• Taking reproducible screenshots of pages or flows
• Filling forms and submitting by clicking buttons
• Finding page elements for scraping or testing
• Multi-page workflows that require programmatic clicks and navigation

Best practices

• Use simple, reliable CSS selectors (element, class, attribute) and test them with browser_find first
• Take screenshots before and after key actions to capture page state for debugging
• Handle connection errors by relaunching the browser and retrying operations
• Break complex tasks into smaller steps to avoid timeouts and verify each step
• When advanced selectors or keyboard input are needed, combine Vibium with cursor-ide-browser or use click-based workarounds

Example use cases

• Collect screenshots from a list of URLs for visual QA or archival
• Automate form data entry across multiple pages and save a filled-form screenshot
• Scrape specific elements (headings, links) by running browser_find and reading returned text/box data
• Automate search workflows: type query, click search button, capture results
• Multi-page flows that navigate, interact, and capture screenshots for documentation