executing-plans

You can execute a development plan with review checkpoints by loading the plan file, tracking tasks iteratively, applying tools to verify changes, and generating a completion report to ensure safe, auditable execution.

The best way to track sub-tasks and verify changes is to automate step-by-step execution within a session, applying built-in review checkpoints at each phase to measure outcomes and prevent misalignment.

Yes, automated plan execution requires isolated workspaces via git worktrees and non-main branch workflows to ensure safe, auditable execution and prevent unreviewed changes from affecting the main codebase.

A subagent handles task management by decomposing complex tasks into sub-tasks, following the plan to deliver measurable outcomes, and applying explicit readiness checks to maintain workflow traceability.

You should avoid using automated stepwise plan execution if you lack an isolated git worktree environment, as explicit readiness checks and non-main branch workflows are strictly required to guarantee safe execution.