subagent-driven-developmentlisted
Install: claude install-skill AidALL/ghost-alice
# Subagent-Driven Development
Dispatch a fresh subagent per task to execute the plan. After every task, run a two-stage review: spec compliance review first, then code quality review.
○ Why subagents
Delegate each task to a specialized agent with an isolated context. Make the instructions and context precise so the agent stays focused on the task and succeeds. The agent must never inherit your session's context or history. You construct exactly what the agent needs. This also preserves your own context for coordination work.
○ Core principle
A fresh subagent per task plus a two-stage review (spec then quality) supports
focused work and faster correction.
## Contents
- [When to Use](#when-to-use)
- [Procedure](#procedure)
- [Model Selection](#model-selection)
- [Handling Implementer States](#handling-implementer-states)
- [Prompt Templates](#prompt-templates)
- [Example Workflow](#example-workflow)
- [Benefits](#benefits)
- [Red Flags](#red-flags)
- [Integration](#integration)
## When to Use
```dot
digraph when_to_use {
"Is there an implementation plan?" [shape=diamond];
"Are the tasks mostly independent?" [shape=diamond];
"Staying in this session?" [shape=diamond];
"subagent-driven-development" [shape=box];
"executing-plans" [shape=box];
"Manual execution or brainstorm first" [shape=box];
"Is there an implementation plan?" -> "Are the tasks mostly independent?" [label="yes"];
"Is there an implementation plan?" -> "Manual execution or