subagent-driven-development

To execute multi-task software plans without context bleed, isolate each task by dispatching a fresh implementer subagent per task. This approach keeps each task focused and prevents cross-task interference during heavyweight code orchestration.

The best way to verify spec compliance and code quality in workflow automation is running two sequential review passes after each task. This validates requirements first, then evaluates code quality, ensuring both spec compliance and high standards.

Yes, you can run continuous code orchestration for mostly independent tasks without unnecessary check-ins. The workflow executes the whole plan continuously, stopping only when it encounters blocked or ambiguous progress that requires controlled continuation rules.

Yes, you need a well-defined implementation plan with mostly independent tasks before starting. This skill fits large, manually invoked plans where tasks require careful orchestration rather than tightly coupled or highly dependent task structures.

Task isolation improves code quality review by using a fresh subagent context for each task, preventing prior task states from skewing assessments. This allows the two-stage review to accurately catch spec mismatches and code-quality issues per task.

You should not use subagents for task planning and code orchestration when tasks are highly dependent on each other or when the implementation plan is poorly defined. This approach requires mostly independent tasks and deliberate orchestration checkpoints.