I would just use a bounding box and then the “Worldposition behind translucency” node. then you would subtract ObjcetPosition from that… then you would do “transform to Z vector” where the (worldposbehindtranslucency-objectpos) is the value to transform and the actor’s forward vector will be the Z.
If you use the matrix3x3transform nodes it will give you the ability to control the yaw as well.