subagent-driven-development

To prevent context pollution during complex development workflows, delegate individual tasks to isolated subagents. This approach keeps the main context clean by having specialized subagents handle specific implementation steps independently.

You can enforce automated code review and spec compliance for parallel tasks by applying a two-stage review loop. This process requires a spec compliance review followed by a code quality review before any subagent task is integrated.

The best way to execute multi-step implementation plans in parallel is to break down the plan into independent sub-tasks. You then spawn fresh subagents to handle each task concurrently, ensuring strict adherence to the original specifications.

To start delegating development tasks to subagents, activate the subagent-driven workflow to execute an implementation plan located in your docs/plans directory. The system will automatically orchestrate the task delegation.

Yes, parallel subagent execution works for independent feature components. You can spawn separate agents for model creation, utility implementation, and endpoint development, verifying each against the original plan before integration.