Guiding virtual fixtures have been proposed as a method for human-robot co-manipulation. They constrain the motion of the robot to task-relevant trajectories, which enables the human to execute the task more efficiently, accurately and/or ergonomically. When sequences of different tasks must be solved, multiple guiding virtual fixtures are required, and the appropriate guide for the current task must be detected automatically. To this end, we propose a novel control scheme for multiple guiding virtual fixtures that are active in parallel. Furthermore, we determine under which conditions using multiple fixtures is stable. Finally, we perform a pilot study for a real-world application with a humanoid robot.