What problem does it solve? Manually testing web UIs and verifying live application behavior is slow and error-prone, and brittle CSS-selector scripts break on every UI change. This Skill drives a real browser through Playwright MCP using text-first accessibility tree snapshots and stable element reference tokens, so agents can navigate, fill forms, inspect network traffic, and verify UI flows without vision models or fragile selectors. ## Core Features & Use Cases - Accessibility-Tree Interaction: Capture page structure with browser_snapshot and click, type, or select elements via [ref=eN] tokens instead of guessing CSS selectors. - Diagnostics & Verification: Inspect console messages, network requests with headers and payloads, and take screenshots to confirm UI state and API contract adherence. - Advanced Scenarios: Manage tabs, dialogs, and mobile viewports, or run custom Playwright code via browser_run_code_unsafe for multi-step assertions. - Use Case: After deploying a new checkout flow locally, navigate to the page, fill the form via batch field population, submit, wait for the confirmation text, and check the network log to verify the order API returned the expected payload. ## Quick Start Use the browser-usage skill to open http://localhost:5173, snapshot the page, fill in the login form, and verify the dashboard loads without console errors.
