Several advanced control laws are available for complex robotic systems such as humanoid robots and mobile manipulators. Controls are usually developed for locomotion or for manipulation purposes. Resulting motions are usually executed sequentially and the potentiality of the robotic platform is not fully exploited. In this work we consider the problem of loco-manipulation planning for a robot with given parametrized control laws known as primitives. Such primitives, may have not been designed to be executed simultaneously and by composing them instability may easily arise. With the proposed approach, primitives combination that guarantee stability of the system are obtained resulting in complex whole-body behavior. A formal definition of motion primitives is provided and a random sampling approach on a manifold with limited dimension is investigated. Probabilistic completeness and asymptotic optimality are also proved. The proposed approach is tested both on a mobile manipulator and on the humanoid robot Walk-Man, performing loco-manipulation tasks.