What problem does it solve? Executing a multi-task implementation plan in a single session often pollutes context, skips reviews, and lets quality drift between tasks. This Skill orchestrates plan execution by dispatching a fresh subagent per task with isolated context, then enforcing a two-stage review (spec compliance first, then code quality) after every task. ## Core Features & Use Cases - Fresh subagent per task: Each task runs in an isolated context with exactly the instructions and scene-setting it needs, preserving the controller's context for coordination. - Two-stage review gates: A spec compliance reviewer verifies nothing is missing or extra, then a code quality reviewer checks maintainability before the task is marked complete. - Status handling and escalation: Implementer subagents report DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, or BLOCKED, with defined recovery paths including model upgrades and task splitting. - Use Case: Given a written plan with five independent tasks, the controller extracts all task text, creates a todo list, dispatches implementer and reviewer subagents per task, and finishes with a final whole-implementation code review. ## Quick Start Use the subagent-driven-development skill to execute the implementation plan in docs/plans/feature-plan.md task by task in this session.
