I also believe that even if one is interested in smooth manifolds, it is much easier and conceptual (or perhaps natural) to define "smooth" manifolds through diffeological spaces, just like the definition of topological manifolds. Actually, we can say that a smooth manifold is a diffeological space that is locally Euclidean: every point in the space has an open neighborhood (with respect to the D-topology) which is diffeomorphic to an open subset of a fixed Euclidean space. After that, one can add Hausdorfness and second-countability requirements, if needed. In this situation, there is no need to talk about the smooth compatibility of charts, because it is automatically verified. Notice that diffeological spaces are an extension of Euclidean spaces in the first place.