subagent-driven-development

To prevent context bleed during multi-task execution, assign each task to a fresh subagent. This orchestrator-driven workflow ensures per-task isolation so previous implementation context does not pollute subsequent tasks.

A two-stage review process evaluates task implementation sequentially. A Spec Reviewer checks requirements compliance first, then a Quality Reviewer assesses code quality, with re-reviews required until the implementation passes both stages.

Orchestrate independent implementation tasks by deploying a coordinated workflow. An orchestrator dispatches tasks to fresh subagents, monitors per-task isolation, and applies staged verification to signal clear completion across the entire plan.

Use role-driven orchestration for plan execution when a plan contains multiple independent tasks fitting in a single session. It minimizes context leakage and accelerates iteration by separating Implementer, Spec Reviewer, and Quality Reviewer roles.

Yes, subagent-driven development is designed for plans with multiple independent tasks that must be completed in a single session. It deploys a fresh subagent per task to maintain isolation while coordinating the overall workflow.