gsd-execute-phase

You can execute phase plans in parallel by using wave-based scheduling, which groups dependent tasks into sequential execution waves and spawns subagents to process each wave concurrently.

Wave-based scheduling is an execution method that divides complex phase plans into dependency-aware groups, allowing isolated subagents to execute tasks in parallel waves for faster overall completion.

You orchestrate subagents by loading phase-scoped context for each wave and assigning isolated subagents to execute each group of plans, ensuring inter-plan dependencies are respected across the workflow.

Parallel phase execution supports checkpoint-aware result aggregation, preserving workflow state during wave-based execution so that completed subagent tasks can be recovered and retried if failures occur.

Subagent orchestration is designed for complex workflows with inter-plan dependencies, using wave-based parallel execution to coordinate isolated context loading and manage task execution across the entire phase.

To run a phase using wave-based parallel execution, you need a defined phase identifier and its associated phase plans, which the orchestrator will automatically group into execution waves and assign to subagents.