I’m sorry but this sounds like a horrible idea from a programming standpoint. Abilities are one thing and in this use-case, since how the input is interpreted is decided by the ability itself and it might be exactly 1-to-1, because there might be as many input interpretations as there are active abilities, the input handling should be inside an ability.
A character should know nothing about it’s abilities other than “It’s active now”.
Implementing some unknown N input types inside a character and then the need to add more in case you add new ability inside your character is a terrible programming practice and I would like to avoid it 