[Question] Implementing multiple actions per turn

[a-ws-m]

Question

I'm trying to implement an API for a card game where each turn has two phases: a drawing phase and then a playing phase. The agent can choose from multiple cards to draw and then can choose a card to play from the resulting hand. One option during the draw phase is to take from a face-down pile, so there may be new observations after the first phase of the turn. Is it possible to implement this kind of a game using Petting Zoo?
I imagine it might be necessary to treat each player as two agents: one responsible for drawing and the other responsible for then playing cards. That seems to be the easiest way to create two different action masks. Of course, in that case the two agents ought to share a reward function, which I'm not sure how to implement.

[a-ws-m]

I'm going with two agents per player, with the first one handling the drawing phase and the second one handling the playing phase. Sharing the reward function seems simple enough. Do let me know if there's a better approach to this please :)