What happens if a reviewer finds an issue?

The implementation subagent is automatically notified of the specific failure (either spec or quality) and must resolve the issue before the task can be marked as complete.

How does this differ from the standard plan execution skill?

Unlike standard execution, this skill operates within the same session for faster iteration and enforces a mandatory two-stage review (spec compliance then quality) for every single task.

What is Subagent-Driven Development?

It is a methodology where a controller agent manages the execution of a plan by delegating individual tasks to fresh subagents, ensuring clean context and high-quality reviews for every step.

Why is a fresh subagent used for every task?

Using fresh subagents prevents 'context pollution,' where information or errors from a previous task might negatively influence the AI's performance on the current task.

Does this skill support Test-Driven Development?

Yes, subagents dispatched through this workflow are instructed to follow TDD practices, implementing tests alongside or before the actual code changes.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: catlab-design

bycatlab-design

0•

Productividad y Flujo de Trabajo

Orchestrates complex implementation plans by dispatching specialized subagents for independent tasks with automated two-stage quality reviews.

Subagent-Driven Development is a high-efficiency workflow designed to execute multi-task implementation plans within a single Claude Code session. By spawning a fresh, independent subagent for every task, the skill prevents context pollution and ensures that each component is built in isolation following TDD principles. The core of this skill is its rigorous two-stage review gate: code is first audited for strict specification compliance, and only after passing is it reviewed for architectural quality, ensuring that the final output is both functional and maintainable.

Características Principales

01Automated task tracking and state management with TodoWrite

02Isolated task execution via fresh subagent dispatching

03Seamless integration with Git worktrees for clean development environments

04Context-aware questioning to resolve ambiguities before implementation

05Two-stage review process: Spec Compliance followed by Code Quality

060 GitHub stars

Casos de Uso

01Executing multi-step feature plans where tasks are modular and independent

02Scaling development velocity in single sessions without manual task handoffs

03Automating high-quality refactoring of complex systems with mandatory review gates

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add catlab-design/realitylauncher-client subagent-driven-development

For use in Claude.ai and ChatGPT

Características Principales

01Automated task tracking and state management with TodoWrite

02Isolated task execution via fresh subagent dispatching

03Seamless integration with Git worktrees for clean development environments

04Context-aware questioning to resolve ambiguities before implementation

05Two-stage review process: Spec Compliance followed by Code Quality

060 GitHub stars

Casos de Uso

01Executing multi-step feature plans where tasks are modular and independent

02Scaling development velocity in single sessions without manual task handoffs

03Automating high-quality refactoring of complex systems with mandatory review gates

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add catlab-design/realitylauncher-client subagent-driven-development

For use in Claude.ai and ChatGPT