What problem does it solve? Many desktop applications (IDEs like Cursor, native Win32 apps, Electron apps) expose no automation API, forcing agents and users to fall back on fragile manual clicking and typing. This Skill gives an agent eyes (screenshots with coordinate grids), a hand (mouse move/click/drag/scroll), a keyboard (CJK/emoji-safe typing, IME-aware key combos), and UI Automation probing to drive any Windows GUI reliably. ## Core Features & Use Cases - Vision & Precision Targeting: Capture full screens, windows, or 1:1 region crops with optional coordinate grid overlays, and get exact control rectangles via UI Automation (ui-find) without spending vision tokens. - Input Automation: Move, click, drag, and scroll the mouse; type text (CJK/emoji/multi-line safe via message or clipboard modes); send key combos atomically with IME detection and switching to avoid silent keystroke loss. - Window & Process Management: Enumerate all top-level windows by handle (multi-window, multi-monitor, negative coordinates supported), focus/maximize/move windows, and wait for windows or pixel stability instead of blind sleeps. - Batching: Run many steps in a single process to eliminate per-call startup cost. - Use Case: Automate a workflow in Cursor IDE — find its window, focus it, send ctrl+shift+p, type a command, verify the result via edit -Mode read, and screenshot the outcome, all in one batch call. ## Quick Start Ask the AI to use the computer-use skill to take a screenshot of the current screen with a coordinate grid, then click a specific button in the target application window.
