What problem does it solve? CDP-based browser automation gets fingerprinted and blocked by sites, runs without your logged-in sessions, and screenshots alone cannot explain why a page broke. This Skill drives the real Chrome/Brave/Helium browser from inside via a Chrome extension, preserving authenticated sessions and passing bot detection, while exposing DOM reads, JS evaluation, console errors, and network traffic to diagnose failures like hydration mismatches and blank pages. ## Core Features & Use Cases - Six browser capability classes: visual capture (DOM-render and pixel screenshots), DOM read (accessibility tree, markdown, refs), JS eval in isolated or main world, passive network capture with HAR/pcapng export, input (click, type, keys, drag), and record/replay of user flows. - macOS Computer Use bridge: drive native apps via the accessibility tree, OS-level trusted input, backgrounded screenshots without focus changes, Apple Events dispatch, Vision OCR, and VM lifecycle management. - Enforced isolation doctrine: a mandatory preflight gate pins every command to a dedicated test Chrome profile and hard-stops rather than ever touching the operator's Default profile; raw screencapture and osascript focus-pulling are forbidden. - Use Case: After deploying a site, verify it visually by capturing a screenshot through the sanctioned Capture.sh wrapper, reading console errors via eval, and exporting the network log as HAR — all inside your signed-in test profile without disturbing your main browser window. ## Quick Start Ask the agent to verify a deployed page by running the preflight isolation gate and then capturing a screenshot of the URL through the Interceptor test profile.
