What problem does it solve? Automating a real, user-visible desktop requires more than HTTP calls: you need to move the mouse naturally, click, type, scroll, and manage windows across native applications on a running X11/KasmVNC display. This Skill provides the operational playbook for driving that desktop the way a person would, while the user watches and can take over. ## Core Features & Use Cases - Desktop Discovery & Toolchain Setup: Measure the true display geometry with xdpyinfo and wmctrl, and bootstrap xdotool, wmctrl, scrot, and imagemagick idempotently. - Human-Like Input: Generate eased, curved mouse paths in Python, plus clicking, dragging, typing, and scrolling via xdotool. - Window Management: List, activate, move, resize, maximize, and tile windows side by side with wmctrl, including the critical fix of removing maximized state before any move/resize. - Screenshot Capture: Capture the full root window with scrot and hand it to a vision-capable reader when the harness cannot render images. - Use Case: A user asks you to open two browsers side by side and compare pages. You measure the root dimensions, launch the windows, remove their maximized state, tile each to half the screen, and capture a screenshot to confirm the layout. ## Quick Start Ask the agent to start the desktop with zipbox-desktop up, then move the mouse to a target and click it while tiling two windows side by side on display :1.
