subagent-driven-development

To orchestrate task execution without context leakage, delegate each task to a fresh subagent with isolated context. This deterministic, plan-driven workflow prevents cross-task interference and preserves session coherence during rapid iterations.

A two-stage review workflow enforces quality by running a spec compliance review first, then a code quality review. This staged validation ensures generated code meets implementation plan requirements before finalizing task dispatch.

Dispatch independent implementation tasks by extracting them from a plan and sending each to a fresh subagent. Predefined prompts for implementers and reviewers ensure consistent context and expectations across all dispatched tasks.

Yes, subagent-driven development is specifically applicable to planning-heavy sessions where tasks are mostly independent or require staged validation. It coordinates per-task dispatch and iterative refinements to ensure rapid, high-quality execution.

You should not use isolated subagents for task dispatch when tasks are highly interdependent and require shared context. The deterministic workflow is designed for independent tasks, preventing cross-task leakage by isolating context.

Yes, subagent-driven development requires predefined prompts for implementers, spec reviewers, and code quality reviewers. These standardized prompts enforce a deterministic workflow and ensure consistent context across all subagents.