Presumably, the model, the bone values and the projection matrix are all passed to the draw calls being examined. You're right that its perhaps easier to eyeba

Presumably, the model, the bone values and the projection matrix are all passed to the draw calls being examined. You're right that its perhaps easier to eyeball the final verts than attempting to reverse-engineer the skinning setup but I would be surprised if it was all that much effort to find the projection matrix and apply the inverse to the final verts. Well, perhaps a process improvement for anyone so inclined to do the digging.