Creating scenarios for training motor skills such as hand-eye coordination using augmented reality (AR) technology has several challenges. One of them, the application of task variability, a principle to foster flexibility and to unanticipated tasks adaptation. Due to its benefits, a model-based approach was applied to implement it in the training scenario development. In this paper, concretization rules were proposed and evaluated, from a task model that describes the practical exercise. The results show the concretization effectiveness. A case study is made for the design of hand-eye coordination exercises, a skill required for surgical procedures such as the transpedicular screws placement in spine surgery.