What problem does it solve? AI coding agents cannot see or interact with live web applications, forcing developers to manually click through pages, verify deployments, and test user flows by hand. ## Core Features & Use Cases - Headless Browser Automation: Navigate URLs, click elements, fill forms, upload files, and handle dialogs through a persistent Chromium session with ~100ms per command. - Ref-Based Interaction: Snapshot the accessibility tree to get stable element references (@e1, @e2) instead of guessing CSS selectors, with diff, annotation, and cursor-interactive scanning modes. - QA Evidence Collection: Take annotated screenshots, check console errors, test responsive layouts, and assert element states to document bugs with reproducible evidence. - Use Case: After pushing a branch, instruct the agent to log into staging, walk through the signup flow, screenshot each page, and report console errors — a full QA pass without opening a browser. ## Quick Start Ask the agent to use the gstack browse skill to open your staging URL, fill in the login form, and screenshot the dashboard to verify the deployment.