This is something I also had on my mind for quite a while but haven't done so far.
I can't remember the exact reason but it should mostly be because there were still
lots of changes happening so keeping them private was better at the time.

What I am thinking of now though, and what would oppose doing this, is that ActionCommands
and ActionFlags are rather some dumb low-level actions that are missing context.
So you can't pass it as PositionalAny as ActionFlags for example wouldn't know if a flag
was already set (unless the flagset was already parsed) or if a value of a flag should be completed.
ActionExecute is mostly a better choice as it includes traverse which handles all of this.

Might still be useful though anyway, so I'll give it some more thought.

Thanks for these explanations.
I had not thought of this PositionalAny problem myself.

Pinging on this to know if there is a reason why the ActionFlags has not been exposed ?

Asking this because the initial intent was to give some way to dynamically complete commands/flags, should consumer libraries need it.

Pinging on this to know if there is a reason why the ActionFlags has not been exposed ?

Asking this because the initial intent was to give some way to dynamically complete commands/flags, should consumer libraries need it.

ActionFlags is a bit more complicated and entangled with with other code. I also have no concrete use case for it in sight at the moment.