The “case-specific” approach depends on the case 
You just need to supply your code with some kind of a reference to the agent being controlled/affected by the task, get it’s controller and extract the brain component from it. The brain component in this case should be a BT component.