That data doesn’t normally exist and is very difficult to get to because skeletal meshes are skinned on the gpu. You have the base mesh and each vertex knows how much influence each bone has over it. Then the skeletal mesh is animated in some way using an anim blueprint and the animations are blended together so the skeleton can be in completely arbitrary positions at any time. Also you can procedurally position the skeleton by just setting bone transforms.
You’d have to run some code to manually transform a vertex based on how it’d deform and get the data on cpu.
This is why you mostly just attach things to sockets or get the bone transforms themselves. You’d probably be better off doing something with that rather than vertices themselves because it’s almost always the better way than getting geometry data directly from an animated mesh