v2.6.6

LearningMachine now injects its context into the Team system prompt, not just the agent path. Teams get the same learned context that individual agents already received, so their behavior reflects it as expected.