How does the two-stage review process work?

After a subagent implements a task, a 'Spec Reviewer' ensures it meets all original requirements. Once passed, a 'Code Quality Reviewer' checks the implementation for best practices and efficiency.

Does this skill work with Test-Driven Development?

Yes, subagents are instructed to follow TDD principles naturally, creating and running tests for each task before submitting it for review.

What is Subagent-Driven Development?

It is a specialized workflow for Claude Code that breaks down a development plan into independent tasks, each handled by a fresh subagent to ensure clean context and high-quality implementation.

Why use subagents instead of a single long session?

Using fresh subagents per task prevents 'context pollution'—the accumulation of irrelevant information that can cause AI models to become confused or hallucinate over long sessions.

When is it better to use this over parallel session execution?

This skill is ideal when you want to stay in your current session for faster iteration and immediate progress tracking without the need for manual handoffs between different environments.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: Cygnusfear

byCygnusfear

0•

Productivity & Workflow

Executes complex development plans by dispatching specialized subagents for task implementation and mandatory multi-stage reviews.

Subagent-Driven Development is a high-precision methodology designed to execute implementation plans within a single Claude Code session. By dispatching a fresh, task-specific subagent for every item in a plan, it eliminates context pollution and ensures rigorous quality control. The skill enforces a mandatory two-stage review process for every task: a specification compliance check followed by a code quality audit. This approach facilitates rapid iteration, maintains high architectural standards through Test-Driven Development (TDD), and ensures that every line of code meets both functional requirements and best practices without requiring constant human-in-the-loop intervention.

Key Features

01Built-in support for Test-Driven Development (TDD) within subagent tasks

02Fresh context subagents for every task to prevent memory and logic pollution

03Same-session execution to avoid the overhead of context switching or handoffs

04Mandatory two-stage review process (Spec Compliance and Code Quality)

05Automated task tracking and progress management via TodoWrite

060 GitHub stars

Use Cases

01Executing multi-task feature implementations from a pre-defined development plan

02Automating complex refactoring projects that require high-precision spec adherence

03Scaling development velocity while maintaining strict code quality and testing standards

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cygnusfear/claude-stuff subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill