subagent-driven-developmentlisted

# Subagent-Driven Development Use this skill to execute an approved implementation plan with isolated subagents. ## Relationship To Context Skills When the work is tracked in `.my-context/tasks/`, use `context-task` as the controller's durable task record. - read the task from `task.md` - use `plan.md` as the current execution source of truth - store task-specific discoveries in `notes.md` - finish with `outcome.md` and completed task state Use `context-memory` when prior durable decisions, architecture notes, or cross-task discoveries may affect the plan or task sequencing. ## Why This Skill Exists The goal is to keep each subagent's context narrow so its decisions stay clear, reusable, and cheaper than carrying one large implementation session. This workflow separates roles: - controller: coordinates the plan and handoffs - `builder`: implements one task - `task-reviewer`: reviews one task before the next begins - `code-reviewer`: performs the broader final code review ## When to Use This Skill Use this skill when: - you already have an approved implementation plan - the plan has multiple tasks or checkpoints - tasks are independent enough to execute sequentially with clean handoffs - you want stronger context isolation between implementation and review Do not use this skill when the work is tiny, highly exploratory, or too tightly coupled to split into task handoffs. ## Core Workflow 1. Read the plan once and identify all tasks 2. Check for obvious contradic