subagent-driven-development

You orchestrate subagent task dispatch by assigning a fresh subagent to each task in your implementation plan, preventing context pollution. This approach enforces a two-stage review workflow for spec compliance and code quality to ensure high quality and fast iteration.

The best way to prevent context pollution is dispatching a fresh subagent per task. This ensures each task starts with a clean slate, enabling parallelizable workflows and maintaining high code quality across multiple independent tasks in a single session.

Two-stage code review works by running a spec compliance review first, followed immediately by a code quality review after each task completes. This sequence ensures the implementation meets original specifications before evaluating the actual code quality.

Yes, you can run subagent-driven development autonomously in headless operation. When invoked by the implement-issue flow, subagents run autonomously with no user prompts, executing tasks and reviews without manual intervention.

You track feature branches by capturing the branch name and using it for all commits throughout the orchestration process. This prevents branch drift and ensures all task commits remain consistently associated with the correct feature branch.

You should use subagent orchestration when you have a structured implementation plan with multiple independent tasks. It is specifically applicable for plans requiring iterative refinement through spec compliance and code quality reviews within a single session.