Human beings perform complex tasks while even keeping balance. Though this is evident for a human, it is very difficult to adapt human motion to humanoid robots, due to the posture redundancy. In this paper we propose a conceptually simple framework of human posture control, scoping in a general way with grasp, task achievement and being on the same time robust to external disturbances. We do this by establishing physically meaningful constraints on the force level. In order to respect all constraints, we solve for a quadratic minimization problem (quadratic programming), yielding the required joint torque controls. Contrarily to most other approaches, we deal also with unilateral contacts (due to friction) as well as bilateral grasps which allow for example for arbitrarily steering or pulling on a handhold. Additionally, and in contrast to classical methods based on ZMP, we account also for contacts not all being in the same plane, as hand-wall contact.