subagent-driven-development

Executing a development plan without context pollution is achieved by dispatching independent subagents for each task. This approach assigns a fresh subagent to every decomposed step, ensuring focused task execution and preventing cross-task interference.

Automating code review and spec compliance is best handled through a two-stage review process after each task execution. This mechanism rigorously checks both spec compliance and code quality automatically, ensuring high-quality code during iterative development sessions.

Task decomposition for implementation plan execution works by automatically breaking down a detailed plan into discrete, manageable tasks. It then assigns each task to an independent subagent for focused execution and subsequent review.

Subagent-driven development is ideal for a feature plan with multiple distinct steps. It ensures each step is implemented correctly and efficiently through automated task decomposition and independent subagents, requiring no manual oversight between tasks.

A detailed plan document is required to use automated subagent task execution. The skill operates by reading an implementation plan, such as one located at a specific markdown path, and breaking it down into individual tasks for the subagents.

You should not use independent subagents for development task execution when your plan lacks multiple distinct steps or requires heavy shared context across tasks. The approach is designed for discrete, manageable tasks that benefit from fresh context to maintain code quality.