executing-plans

Batch execution of a development plan breaks tasks into small, defined groups, defaulting to three, and enforces mandatory review checkpoints after each cycle to pause for human feedback before proceeding.

Managing AI task completion during code implementation is achieved through iterative batch processing that tracks task status, enforces reporting at checkpoints, and halts execution immediately upon encountering blockers.

Error handling during plan execution stops the batch processing immediately upon encountering blockers, providing clear guidance to ask for help and preventing the AI from proceeding without human oversight.

A detailed written implementation plan is required as input, because task execution processes predefined bite-sized steps sequentially, managing completion and verification for architect review rather than generating the plan.

Iterative development workflows for plan execution integrate directly with git worktrees and plan writing skills, providing a complete development lifecycle that manages task completion within isolated branch environments.