I don't know if Torque 3D supports ragdoll and bones bouding box collisions, but you could go further and check each bones volume box visibility.
I think Steve posted a resource for getting the player node positions and transform, not long ago. Maybe it was in c++ subforum.I would suggest expanding the engine slightly as to allow you to easily get the transformed positions of nodes in the player mesh. I did this to approximate bullet collisions on players' limbs with really good success. In this way you can do ray casts to specific points on the animated skeleton (hand, elbow, knees, etc) to see if they're poking out of cover. I'm several engine iterations out of date now so I doubt my old code would be much use, but I'm sure you could poke around at how mount points are used as a jump start.