Disclaimer: I don’t know what you know or your experience, or your priorities (cinematics vs gameplay) so forgive me if this is too basic.
Depends if ‘lightweight’ means in maya or in engine perf cost. In maya use matrix nodes not constraints, perhaps. In engine always consider how big it is on screen and construct your joint /mesh /texture fidelity appropriately.
again ,you need clarify: in maya or in engine. My career has always been building and using in house tools. For how to think about game rigging,
I recommend David Hunt’s 2009 GDC presentation on Modular Procedural Rigging
Also, build tools that allow you to quickly ingest any change that comes in. Tear down and rebuild a rig at the touch of a button. Have a process to save, reload, a transfer skin weights. NG Skintools is essential, imho
Our engine does not use them. Nor have I seen one that does. In an engine you get joint deformation primarily. We did recently implement wrinkle map blending for cinematics. Maybe Unreal or Unity has corrective blends, idk. In any case: Cinematic cut scenes only. I would never consider it for gameplay.
Be surgical about what you export. I know Unity looks like it can use Maya or Max files directly, but It runs the FBX API under the hood and converts it into its native mesh/skinMesh representation. Leaving a Maya file in your asset tree will bloat things unnecessarily. And speaking of native mesh/skinMesh representations: whatever engine you use, you should read the API docs on those native representations. “If you don’t understand the data, you don’t understand the problem” as Mike Acton said. Read up on Game Animation Programming and know what is going on, do not tolerate magic black boxes in your understanding.