My initial reaction is to say I can't imagine a way to avoid the frustration you mentioned. You might be able to hard-code a bunch of specific scenarios of commands to try to fix an issue, but then it would just be like a "troubleshooting dashboard" - not really managing the environment. I mean, think about how easy it is to fat-finger something on a touch screen. Maybe I'm just too clumsy, but I wouldn't want to take that chance on a production server.