applying-effort-policy

Resolving the correct Claude model and effort level for spawned sessions involves applying a policy overlay that checks spawn-time overrides, role templates, and task-type precedence to return concrete configuration values. This ensures orchestrators launch child sessions with the exact tier settings required.

When the effort policy JSON overlay is missing or invalid during session spawning, the resolution process validates the JSON structure and automatically falls back to the doctrine baseline configuration. This guarantees spawned sessions still receive valid model and effort settings without manual intervention.

Role templates and task-type overrides interact through a strict precedence hierarchy where spawn-time overrides take priority, followed by role-specific overrides, task-type overrides, and finally the baseline doctrine. The resolution applies this cascade to yield the effective model and effort values for each session.

Effort policy resolution works across both plugin canonical and consumer mirror policy files by reading the JSON overlay from either source. It validates the structure and applies the appropriate precedence rules so spawned sessions use consistent configurations regardless of the file origin.

Ensuring consistent Claude model settings across orchestrator child-spawn commands requires resolving the effective effort policy before rendering templates or launching agents. The resolution validates overlays, honors precedence, and supplies concrete values so phase, epic, sprint, and task agents all inherit correct tier configurations.

A valid JSON overlay is not strictly required to spawn sprint or task agents because the resolution tolerates missing or invalid files by falling back to the doctrine baseline. However, providing a valid overlay lets you define custom role and task-type overrides for precise effort control.