Action and responsibility are often linked to control, but skilled behavior reveals a tension in the way we think about control: control is sometimes thought to require flexibility, other times reliability. Yet, flexibility and reliability compete with one another, as we see in skilled behavior; the habituation behind skill provides more reliability, but less flexibility. Thus, accounts of intentional action are split on how to handle skilled behavior. I call this the “problem of skilled behavior” – skilled behavior seems both more and less within our control, so how can we account for it within a unitary model of control? Most traditional accounts adopt a unitary model of control within a "dual-process" model of behavior, which treats all behavior as either controlled or automatic, or some combination of these. I suggest instead a "hierarchical model" of behavioral control, in which behavior can benefit from either attention-based control, which has the benefit of flexibility, or strategic automaticity, which has the benefit of reliability. I argue that action and responsibility can be based in either form of control against "attention-for-action" theorists, such as Fridland and Wu.