subagent-driven-development

Automate plan execution by dispatching a fresh implementer subagent per task to prevent context leakage. Each isolated subagent executes its assigned task independently, enabling parallel progress across loosely coupled workflows.

Per-task code review works by running a task-review immediately after each subagent finishes implementation. This structured review workflow enforces spec compliance and checks code quality before moving to the next task.

Maintaining a progress ledger across multiple tasks requires a structured workflow that tracks isolated subagent execution and per-task reviews. This orchestration ensures coherent plan execution and end-to-end code quality.

Subagent-driven development applies to plan-driven work where tasks are largely independent or loosely coupled. Tightly coupled tasks requiring shared context are not ideal for this isolated subagent execution approach.

Ensure end-to-end quality by running a final whole-branch review after all independent subagents complete. This broad aggregation pass consolidates changes across tasks to verify coherence and overall code quality.

Human-in-loop interruptions are not needed during AI-assistant task execution unless the subagent is blocked. Guided execution constraints enforce minimal narration between tool calls to proceed autonomously.