What problem does it solve? Executing a multi-task implementation plan in a single AI session pollutes context, skips quality gates, and loses progress after compaction. This Skill orchestrates plan execution by delegating each task to an isolated subagent with curated context, enforcing review after every task, and tracking progress in a durable ledger. ## Core Features & Use Cases - Per-task subagent dispatch: Each task gets a fresh implementer subagent with a file-based task brief, explicit constraints, and a report contract, keeping the controller's context clean. - Two-stage review gates: A task reviewer verifies spec compliance and code quality per task, followed by a broad whole-branch review at the end, with fix subagents for Critical and Important findings. - Durable progress tracking: A plan-scoped ledger and review-package scripts record commits, snapshot tree SHAs, and verification evidence so work resumes correctly after session compaction. - Use Case: Given a written implementation plan with eight independent tasks, the controller extracts each task brief, dispatches implementers sequentially, reviews each diff, and finishes with a whole-branch code review before merge. ## Quick Start Use subagent-driven development to execute the implementation plan at docs/plans/feature-plan.md task by task with reviews.
