subagent-driven-development

Parallel subagent orchestration spawns dedicated subagents to execute independent plan tasks concurrently in isolated swarms. It enables concurrent work while ensuring spec compliance and code-quality reviews through a centralized coordinator.

To set up parallel task execution, instruct the system to deploy the plan using a bones-powered coordinator. The coordinator manages task dispatch, assigns per-task worktrees, and handles re-dispatch flows for isolated subagent swarms.

The two-stage review process executes a spec compliance check followed by a code quality review for each completed task. This ensures correctness before merging concurrent work back from the isolated subagent swarms.

Use parallel subagents for plans containing independent tasks that benefit from concurrent execution. If tasks have heavy dependencies requiring sequential completion, isolated parallel swarms will not provide execution advantages.

Yes, you can coordinate bones task graphs by reporting progress back to the coordinator. The system manages task dispatch and heartbeat updates, ensuring the parallel subagents remain synchronized with the overall plan scope.

Orchestrating isolated subagent swarms is limited by task dependencies; heavily interdependent tasks negate parallel execution benefits. Additionally, managing concurrent worktrees requires coordinator overhead for dispatch and two-stage reviews.