subagent-driven-development

Subagent orchestration dispatches a fresh subagent per task with isolated context to execute distributed work safely. It manages two-stage reviews to ensure spec compliance and code quality after each task completes.

To prevent context leakage in parallel coding tasks, assign each task to an independent subagent with an isolated context. This ensures safe, parallelizable execution without context bleed between concurrent operations.

Two-stage code review for plan management tasks runs a spec compliance review first, followed by a code quality review. This sequence ensures the implementation meets requirements before evaluating code standards.

Yes, you can use prompt engineering to define implementer, spec-reviewer, and code-quality-reviewer prompts. These distinct roles enforce a disciplined workflow that guarantees spec compliance and code quality.

Use isolated subagents for code review and plan execution when tasks in a session are largely independent. This approach enables fast, quality-driven iteration by preventing context leakage across self-contained task cycles.

Subagent-driven development requires no external dependencies for task orchestration. It implements a self-contained workflow that manages task cycles from plan to merge-ready changes using internal reviewer prompts.