gsd-execute-phase

You can execute phase plans in parallel by using wave-based parallelization, which groups plans into dependency-aware waves and runs them concurrently to reduce overall runtime.

Wave scheduling is an orchestration method that analyzes plan dependencies and routes execution into sequential waves, allowing subagents to perform plan-level work in parallel within each wave.

You can filter parallel execution by providing specific tokens as arguments to run only certain wave numbers or to trigger gap-only execution for remaining dependency-heavy plan sets.

Yes, agent subtasking supports dependency-heavy GitHub workflows by spawning subagents that load the full execute-plan context while a lean orchestrator preserves checkpoint and verification gates.

Yes, you must run the end-to-end execute-phase workflow to preserve wave, checkpoint, verification, and state gates while orchestrating dependency-aware parallel execution across subagents.

You should avoid wave-based parallelization for small or independent plan sets where dependency analysis overhead outweighs runtime benefits, or when interactive sequential execution is required.