Why are there two review stages?

The first stage ensures the code meets the specific technical requirements (spec), while the second focuses on maintainability, style, and architectural patterns (quality).

How does this differ from standard plan execution?

Unlike parallel session execution, this stays in the same session to reduce context-switch overhead while maintaining task isolation through fresh subagents.

Can I use this for tasks that are tightly coupled?

This skill is most effective for independent tasks. If tasks are tightly coupled, manual execution or deeper analysis is recommended to prevent conflicts.

What is Subagent-Driven Development?

It is a methodology where a controller agent dispatches specialized subagents to implement tasks and perform multi-stage reviews to ensure high-quality code within a single session.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: cloudfieldcz

bycloudfieldcz

•

Productivity & Workflow

Executes multi-task implementation plans using specialized subagents for development and rigorous two-stage reviews within a single session.

Subagent-Driven Development is a high-efficiency workflow designed to execute implementation plans by delegating individual tasks to dedicated subagents. By maintaining the current session while spawning 'fresh' subagents for specific tasks, this skill prevents context pollution and ensures high-quality output. The workflow incorporates a mandatory two-stage review process—spec compliance first, followed by code quality—ensuring that every task not only meets requirements but also adheres to best practices before moving to the next item in the plan.

Key Features

01Mandatory two-stage review process (Spec Compliance and Code Quality)

02Efficient same-session execution to minimize context-switching overhead

03Context-aware task extraction from implementation plans

042 GitHub stars

05Automated dispatching of specialized subagents for implementation and review

06Integrated Test-Driven Development (TDD) enforcement for each task

Use Cases

01Scaling development speed by automating the review loop for routine tasks

02Implementing complex refactoring plans requiring iterative validation

03Executing multi-step feature roadmaps with independent components

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cloudfieldcz/cf-powers subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill