The design of a compact and effective representation is a critical step in fully utilizing the abundant raw mesh‐based animation resources. In this paper, we present a high‐level control structure for animated mesh sequences referred to as an adaptive skeleton‐driven cage representation. This approach combines a skeleton and cage to express time‐varying shape details and offer flexible control for rigid limb motions. The initial skeleton is inferred by sketching on the rest pose mesh. The corresponding cage is constructed by using an adaptive cross‐section‐based method. Then, the initial skeleton and cage are propagated to the other meshes in the sequence. Additionally, the generated cage sequence can be automatically refined to improve the mesh reconstruction quality. Our results and comparisons demonstrate that the proposed approach is an intuitive, compact, and efficient representation. We demonstrate its potential through a variety of applications, such as animation compression, deformation transfer and pose editing. Copyright © 2014 John Wiley & Sons, Ltd.