The window had checkboxes corresponding to the options and text fields corresponding to their values, if any. I can't remember if it parsed the output of --help or equivalent, though.

It was all driven by data custom-created for each command, it wasn't done by parsing -h (no --help in those days) or man pages. That would've been some combination of research project and nightmare. (It also wasn't done the VMS way, by having a nice set of libraries for defining commands and their arguments and flags and how to parse them.)

I also don't recall if such commands were supported in pipelines, though I see no reason why they should not have been.

At least in Macintosh Programmer's Workshop, what Commando produced for you was the command line invocation, not the actual running process. So you'd type a command, invoke Commando, and when you clicked OK the command line you were entering would be updated with what you'd selected in the dialog box.