subagent-driven-development

Multi-task implementation plans are coordinated by parsing the plan into a checklist, delegating each task to a fresh subagent with full context, and enforcing two-stage reviews before final integration.

Automated code review for delegated tasks uses a two-stage process: first verifying spec compliance, then checking code quality before marking the task complete and moving to integration.

Large refactors are broken down by reading the implementation plan once, extracting every task into a clear execution checklist, and assigning each isolated task to a new subagent.

Yes, subagent delegation suits independent feature work and validation-heavy refactors by executing each task separately through fresh subagents with mandatory spec compliance and quality reviews.

Fresh subagent reviews are needed for multi-step software features and independent refactors where correctness, consistency, and review discipline matter across each separately executed task.

This approach requires a completed implementation plan upfront and depends on per-task delegation and multi-stage reviews, making it less suited for single-task changes or workflows lacking clear task definitions.