subagent-driven-development

Automating implementation plan execution involves dispatching independent tasks to fresh subagents, applying spec compliance and code quality reviews, and finishing branches via git worktrees to prevent context pollution and manual coordination bottlenecks.

Subagent-driven development is an execution pattern where each task in an implementation plan runs in an isolated agent, followed by a two-stage review process to ensure spec compliance and code quality without manual hand-offs or context leakage.

Automatic code reviews are triggered by running a two-stage review process after each task completes, first checking spec compliance then evaluating code quality, ensuring every implementation piece meets specifications and quality standards automatically.

Git worktrees are integrated directly into the subagent workflow to isolate task execution and coordinate final branch finishing, allowing independent implementation pieces to be developed and reviewed without sharing context or creating bottlenecks.

Fresh subagents should be avoided when tasks are highly interdependent and require sharing context across execution boundaries, as this approach is designed specifically for mostly independent tasks that can be completed in isolated agent environments.

Coordinating spec compliance and code quality reviews is best handled by a two-stage automated pipeline that sequentially validates specifications first, then checks code quality for every task dispatched to a fresh subagent.