Enabling robots to quickly learn manipulation skills is an important, yet\nchallenging problem. Such manipulation skills should be flexible, e.g., be able\nadapt to the current workspace configuration. Furthermore, to accomplish\ncomplex manipulation tasks, robots should be able to sequence several skills\nand adapt them to changing situations. In this work, we propose a rapid robot\nskill-sequencing algorithm, where the skills are encoded by object-centric\nhidden semi-Markov models. The learned skill models can encode multimodal\n(temporal and spatial) trajectory distributions. This approach significantly\nreduces manual modeling efforts, while ensuring a high degree of flexibility\nand re-usability of learned skills. Given a task goal and a set of generic\nskills, our framework computes smooth transitions between skill instances. To\ncompute the corresponding optimal end-effector trajectory in task space we rely\non Riemannian optimal controller. We demonstrate this approach on a 7 DoF robot\narm for industrial assembly tasks.\n