wtf is tart-skills?

This project gives an AI agent running on Linux a way to control a full macOS desktop, even though the agent has no Apple hardware of its own. It does this through Tart, a tool that runs macOS virtual machines on Apple Silicon Macs. Since Tart only works on a real Mac, the agent cannot use it directly, so this project bridges the gap with two SSH connections: one from the Linux agent to a Mac host, and a second from that Mac host into the macOS virtual machine itself. Once connected, the agent can take screenshots of the virtual machine's screen, click and type inside it, launch the IntelliJ IDEA code editor, open a project, and even record what happens as a video file. This effectively gives the agent eyes and hands inside a real macOS graphical session, Finder, Dock, wallpaper and all, so it can test how software actually looks and behaves on a Mac. The project is organized as a set of individual skills, each one handling a specific job: setting up the initial connection, creating and managing the virtual machine, taking screenshots, launching and driving IntelliJ IDEA, recording video, and caching IDE software so it does not need to be downloaded again for every new virtual machine. Because the Mac host may be shared across many tasks at once, each task is expected to create its own uniquely named virtual machine and clean it up when finished, so tasks do not interfere with each other. To use this, you need an Apple Silicon Mac (M1 or newer) with Tart installed, SSH remote login turned on, and SSH access set up so commands can run without typing a password each time. On the Linux side, only a standard SSH client is required. Everything the virtual machine itself needs, like screen automation tools and video recording, gets installed automatically by one of the setup commands. The project is written primarily in Shell scripts.