Provide mechanism to query model/parser for list of valid commands
Jon Ingold's work with Glk and type-ahead command hints is wonderful, but it's also a very complex piece of code.
I believe that this is a significant path to follow outside of the user interface and strictly within the game engine.
My suggestion to enable the author the ability to list "potential" commands, sort them on some probability of success, and let the author review them programmatically.
David Cornelson commented
To clarify further...the idea is to have the game engine look at all in-scope scenery, objects, and commands and attempt (but not execute) each statement. The result would be the command and a rating where the rating is success, failure, undetermined. We might also expect a third result for each command telling is it results in a default or author-specified result.
The benefits would be sizable. The IDE could use this to help the author identify and implement responses. This could lead to a straight-forward way of mass-editing responses.
Of course we could also use this to offer type-ahead facilities like Jon Ingold's new extension, or as a way to convey command hints to the game player.