What problem does it solve? Executing a multi-task implementation plan in a single AI session pollutes context, loses progress after compaction, and ships unreviewed code. This Skill orchestrates plan execution through isolated subagents so each task gets fresh context, independent review, and a persistent ledger that survives session interruptions. ## Core Features & Use Cases - Fresh subagent per task: Dispatches an implementer subagent with a task brief file, report file, and explicit model selection, keeping the controller's context clean for coordination. - Two-stage review gates: Runs a task-scoped review (spec compliance plus code quality) after every task and a broad whole-branch review at the end, with a bounded five-round fix loop and adjudication rules. - Ledger-based recovery: Tracks progress in a per-plan workspace ledger and generates review packages via bundled scripts, so work resumes correctly after context compaction. - Use Case: Given a written implementation plan with ten tasks, the controller dispatches implementers one at a time, reviews each diff, loops on findings, and finishes with a final whole-branch review before merging. ## Quick Start Ask the AI to execute your implementation plan using subagent-driven development, pointing it at your plan file so it can dispatch implementer and reviewer subagents per task.
