There's more...

We could extend this approach by mixing it with weighted blending, so we would have a more robust architecture, achieving extra precision in the way the behaviors impact on the agent at every priority level:

foreach (Steering singleSteering in group) 
{ 
    steering.linear += singleSteering.linear * weight; 
    steering.angular += singleSteering.angular * weight; 
}