subagent-driven-development

To orchestrate complex implementation plans using subagents, you dispatch fresh, independent subagents to execute individual plan tasks. This approach preserves context clarity and avoids cross-task contamination by handling each task in isolation.

The two-stage review process runs after each task is executed by a subagent. It first conducts a spec compliance review to ensure alignment with requirements, then follows with a code quality review to enforce quality standards.

Yes, subagent-driven development works across sessions with partially independent tasks. It provides end-to-end orchestration including plan extraction, per-task handoffs, and automatic progress tracking to maintain a structured development flow.

You handle in-task branching by utilizing the subagent dispatching framework, which supports in-task branching alongside self-review prompts and structured reviewer prompts to manage complex task execution and automated review loops.

The best way to speed up feature development and reduce context switching is delegating tasks to fresh subagents. This method enables rapid iteration through automated review loops and built-in QA gates, accelerating single-session environments.