Personally, I would first try to get the core functionality right, that is, finding windows and sending events to them, instead of worrying about how to make things look pretty.

For X Windows, there is X11::GUITest, which, if it is anything like its Win32 counterpart, can send keystrokes and mouseclicks to windows. I'd first write a self-contained program that sends a series of keystrokes to a given window. Making a pretty interface will still take a lot of time, but if you don't have the core functionality, all the chrome of a shiny GUI won't help you make your application.