execute-tasks

You orchestrate Claude Code tasks to run in parallel by executing them in dependency-aware waves. This approach launches parallel agents simultaneously while respecting task dependencies to reduce manual coordination and idle time.

Wave-based task execution is an orchestration mechanism that groups tasks into execution waves based on their dependencies. It resolves dependencies automatically, launching unblocked tasks concurrently before merging their results across subsequent waves.

You configure max_parallel and retries for autonomous task workflows by setting these required parameters before execution. Max_parallel limits concurrent agent launches per wave, while retries define adaptive failure thresholds for resilient session state tracking.

Yes, Claude Code task execution supports per-task context isolation. It isolates context for each parallel agent during wave execution and later merges the results, enabling independent retries and seamless knowledge sharing across the workflow.

Yes, you can filter specific task groups during an autonomous execution run. The orchestration mechanism supports optional task-group filtering, allowing you to target specific subsets of spec-generated or general tasks while maintaining session lifecycle management.