subagent-driven-development

Subagent orchestration prevents context bleed by spawning a fresh subagent for each independent task, keeping work focused and isolated. It enforces compliance through explicit gate-style checks and per-task context provisioning.

Two-stage code review validates implementation by first checking specification compliance, then assessing code quality. This structured process catches issues early by verifying the code against the plan before evaluating its overall quality.

Automate task delegation by dispatching a new implementer subagent for each task from the plan. You complete a two-stage review cycle until the task passes both spec and quality checks, ensuring rapid iteration with rigorous validation.

Use fresh subagents for cross-task execution when your plan involves multiple independent tasks where rapid iteration and rigorous validation are critical. The process spans end-to-end from plan extraction to a final integration review.

Isolated subagents are limited to plans with multiple independent tasks. Since each subagent is fresh, they rely entirely on explicit per-task context provisioning, making them unsuitable for highly interdependent tasks requiring shared evolving context.

The best way to ensure spec compliance is by enforcing structured gate-style checks during a two-stage review. The first stage explicitly verifies implementation against the plan before proceeding to the final integration review.