Hi,
In the coordinate system of a depth camera, local X would have coordinates (orient.matrix[0], orient.matrix[3], orient.matrix[6]), local Y would have coordinates (orient.matrix[1], orient.matrix[4], orient.matrix[7]), and local Z would have coordinates (orient.matrix[2], orient.matrix[5], orient.matrix[8]).
As for the question about the T-Pose, please take a look at this video (red vector is X, green is Y, blue is Z): https://www.dropbox.com/s/3boo4tr2a1e7umw/Nuitrack_Joint_Orientation.mp4?dl=0