Thanks, would really depend on the problem to be solved and also your end users\clients I would say. Both are good in their own ways, I like the PDPs from SHAP and even the explainers, while Skater has a lot of model-agnostic capabilities like bayesian rule lists, surrogate trees and they are adding more stuff over time which can be pretty useful too. Using what works best for you and going ahead with that would be my suggestion!