One of the most remarkable feats of our brain is the ability to learn a sheer endless number of motor skills. This capacity depends on a distributed motor network, and while many of its components have been identified, less is known about their specific roles and interactions. We probe this network through the lens of complex, spatiotemporally precise movement patterns we train in rats. We identified the dorsolateral striatum (DLS), the main motor-related input nucleus of the basal ganglia, as a critical hub: Neuronal activity in the DLS encodes the detailed kinematics of the learned motor skills and loss of the DLS disrupts skill execution. During learning, the DLS requires tutoring by motor cortical inputs, which likely includes gating of plasticity at thalamic input synapses in the DLS. While these thalamic inputs remain necessary for skill execution, the motor cortical inputs become dispensable. Together, our results show that the basal ganglia can play a role in controlling complex learned behaviors beyond traditional models and that their cortical and thalamic inputs play novel and distinct functions in skill learning and execution. Furthermore, these findings provide a starting point for our investigations into multi-skill learning, probing how the brain solves the challenge of forming and storing countless motor skills using the same neuronal substrates and circuits.