wombatoperator/omarchy-voice ? reverse-engineered prompt

Build me a voice control app for my Omarchy desktop that lets me ask OMA to move windows, open apps, switch workspaces, read the current page, and manage background tasks. It should use OpenAI for speech and planning, then use local tools to actually do the desktop actions through a safety check so risky commands need confirmation. I want it to start muted by default, with a simple way to toggle listening, confirm actions, cancel them, and check status. It should also be able to inspect the camera on demand, with a local preview and support for vision models when needed. Please include a doctor style check for missing desktop tools, a clean install and uninstall flow, and sensible defaults for Omarchy on Linux. If you need current setup details or package behavior, look up the latest docs online.