In this paper we present some preliminary results regarding a
system for developing autonomous agents that combine
reactivity to an uncertain and rapidly changing environment
and  commitment to prespecified tactics involving cooperation.
A central requirement is that the behavior specification should be done by
people who are not computer and AI specialists.
The original application was air-combat simulation.
Here we consider the simulated soccer domain in perspective of the RoboCup
competition.
The behavior of an agent is specified in terms of prioritized rules
organized into a decision tree. 
Coordinated behaviors are encoded in the decision trees of the individual
agents. 
What link individual behavior descriptions together
are explicit communication and common means to describe the situation
the agents are in.