What problem does it solve?
Coordinating complex implementation plans across multiple independent tasks often leads to context pollution, skipped reviews, and uneven quality. This Skill provides a disciplined, repeatable workflow that uses fresh subagents per task and enforces a two-stage review (spec compliance first, then code quality) to ensure consistent, high-quality delivery.
Core Features & Use Cases
- Per-Task Subagents: Spawn a clean implementer subagent for each task to avoid context leakage and accelerate iteration.
- Two-Stage Reviews: Enforce spec compliance before code quality review, with re-review loops until approval.
- Controller-Oriented Orchestration: Extract all tasks upfront, provide complete context to subagents, and record completion, enabling smooth, same-session execution of plans. A common use case is delivering a multi-step feature plan where tasks are mostly independent (e.g., add an install command, then implement recovery modes), each passing spec and quality gates before moving on.
Quick Start
Use subagent-driven-development to execute the current plan by dispatching an implementer for Task 1 with full task text and context, then run spec review followed by code quality review, repeating for each task until all are approved.
