It may be a good idea to wrap this variable in a function as well for future refactoring.

We call this a universal-argument, so I would avoid adding a new term, prefix-arg.
How about the name *universal-argument* and (universal-argument-of-last-command) ?
If you have any other good names, please let me know.

Thanks for your comment!
Yeah, "naming" is one of the reasons I wanted to ask you because we don't need to have the same name as Emacs has.
I'll change the variable to have universal-argument in its name.

It may be a good idea to wrap this variable in a function as well for future refactoring.

Actually, I'm not sure whether it should be wrapped with a function.
If it's exposed as a special variable, it could be convenient when people want it to be changed in a lexical scope, like:

(let ((*universal-argument* 10))
  (call-command 'something nil))

To be clear, I'm not confident with it, so I'd like to hear if there's some kind of rule (or habit) in Lem to do, or if you prefer it for some reason.

It could be a reason to wrap by a function that the variable is used differently in some contexts.
For example, as you mentioned, universal-argument-of-last-command implies there's also this-universal-argument or so.

If it's exposed as a special variable, it could be convenient when people want it to be changed in a lexical scope, like:

In this case, the second argument of call-command is universal-argument, so binding of the special variable appears unnecessary. (but with a little more complex code, it could certainly be useful.)

Wrapping a function makes it easier to insert breaks at the point of call, and I believe it increases maintainability.

Alright, at least in my use case, it's not necessary to be a special variable, so I'll go with a function style.