Another option would be to create an additional state with a name similar to BeforeEnterSimulation. The agents can be spawned at model start, and they sit in this state until the time specified which sould have been read as a parameter and thus can be used on the state transition.