@Anixan_Skystalker, I think I am following what you are seeing, but if I get anything wrong, please ask more questions.
From what I understand, you are seeing a couple of issues which are that you are not seeing an anchor point transform when importing the mesh and that the anchor point no longer follows the head when the head animates which appears that the anchor is no longer parented to the head mesh.
If this is correct, I believe that both of these issues are coming from the same root condition and, believe it or not, are acting as they should. To explain what is happening is a little abstract, so please bear with me and ask questions if I leave you confused.
When you skin a mesh to a skeleton, the vertices of the mesh follows the bones within the skeleton based on two things: the skinning weight to any given bone and the offset from that bone derived from the bind pose of the mesh. However, if you were to look at the bounding box of a mesh being deformed by the skeleton in Babylon.js, you will see that the bounding box does not move. Note in the gif below I have the bounding boxes displayed for the body and dome meshes while the animation is playing.
This is because the bounding box is computed before the skinning matrix is applied to the mesh. So if you are to parent your anchor to the mesh, it will inherit the position of the mesh, which is shown by the bounding box, before the skinning matrix is applied and would appear not to move with the skinned mesh.
The better way to do this is to parent your attachments to the bones in your skeleton since that is where the motion is derived from and is not affected by the lack of skinning matrix to compute the final position. But this also comes with some challenges if you have to also factor in an offset from the bone to get your attachment in the correct position. For example, if you want to put a hat on a character, the head for the bone may not be located where you would want the root position of the hat to be so you will need an offset.
There are two ways to do this and the first would be a slight modification of what you started with. Instead of creating a cube for your attachment position, use an empty in Blender. This still gives you a transform, but you don’t have any mesh associated with that transform so you don’t need to worry about more vertices or disabling the render once you are in engine. Then you will want to set the empty as a child of the bone driving the part of the mesh you want to attach to. If you want your attachment point for a hat, you would set your empty as a child of the bone driving the head and then set the position of the empty to be the attachment point of the hat. This offset will remain constant and the empty will move with the parent bone.
The second way is to create extra bones within your skeleton that are children of the bones driving the attachment point and position the bones in the location you want your attachment point to be. Again the joint offset will remain and then you want to set the parent of your attachment to be the attachment bone.
Some considerations on this approach would be to make sure that the orientation of the bones used for attachment align with the expected orientation of the attached mesh. If the orientation is wrong, the initial rotation of your attached mesh will be wrong. You will also want to make sure that no vertices have any weight influence being driven by the attachment bones. This could cause issues with the deformation of the mesh and simply giving these bones zero weight would be best.
There are drawbacks to both paths, however, and you need to choose the path that best suits your end goals. For empties as attachment points being children of the bones, you will end up with more transforms in your scene, which does have a compute cost. But this is a really flexible system because you can easily go back and add more attachment locations in the future if you see the system expanding. Using extra bones at attachment points is cheaper in the sense that you aren’t adding more transforms to the scene but may be less ideal to expand in the future. You may find that you need to reskin your meshes to add more attachment bones in the future. However, if you don’t plan to do it often, or if you are willing to take that maintenance on it may be the better option if you have a lot of attachments on your characters.
In terms of the transform node not showing on your cube when looking at it in the inspector, I believe this is due to the fact that your cube is not skinned to the skeleton. I believe we only display the transform of the mesh separately like this when the mesh is skinned.
I hope this helps, but please ping me back if anything isn’t clear.