Prevent configuration drift by scoping team model inheritance

Child agents now inherit only the primary model from their parent. Auxiliary models for output, parsing, or reasoning are no longer inherited, ensuring predictable configurations and reducing hidden coupling across teams of agents. If you relied on the previous behavior, explicitly set auxiliary models on the child.

Details

Primary model inheritance remains; auxiliary model inheritance is removed
Improves correctness and clarity in multi-agent team configurations
Review configs if you previously depended on implicit auxiliary inheritance

Who this is for: Teams composing agent hierarchies or reusable team templates that need reliable, isolated model configurations.