This is done by design. So that a subtree regardless of where it is used, it can be parametrized based on the agent using it, since all blackboard variables are propagated down to all subtrees. This way, you only have one place to look for all the variables and that is the single blackboad that the agent has.
There was a time when I tried allowing a seperate blackboard to each graph, and in the end it really became confusin so I didn’t continue on that path.
Gavalakis
