How does this differ from standard plan execution?

Unlike standard execution, this skill stays in the same session while spawning fresh subagents for each task, ensuring no context pollution while maintaining mandatory two-stage review gates.

What is the benefit of a two-stage review?

The first stage ensures the subagent actually met the requirements (Spec Compliance), and the second stage ensures the code is maintainable and performant (Code Quality), preventing over-engineering or logic gaps.

Do I need an implementation plan before using this skill?

Yes, this skill is specifically designed to execute an existing plan (e.g., created by superpowers:writing-plans) that contains independent tasks.

Does the implementer subagent fix its own mistakes?

Yes. If a reviewer identifies issues, the implementer subagent is re-engaged to fix them, and the review process repeats until both gates are cleared.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: tianzecn

bytianzecn

•

Productivity & Workflow

Executes complex implementation plans by dispatching specialized subagents with integrated multi-stage review gates.

This skill provides a structured workflow for Claude to execute multi-task implementation plans within a single session. By dispatching a fresh implementer subagent for each independent task, it prevents context pollution and maintains high velocity. The core of the process is a rigorous two-stage review system: first, a spec-reviewer ensures full compliance with requirements, followed by a code-quality-reviewer who validates architecture and style. This systematic approach allows for fast, high-quality iteration without the need for manual human-in-the-loop oversight between individual sub-tasks.

Key Features

01Task progress tracking via TodoWrite integration

021 GitHub stars

03Automated context management to prevent session drift

04Integrated Test-Driven Development (TDD) and self-review cycles

05Two-stage review process covering spec compliance and code quality

06Specialized subagent dispatch for isolated task execution

Use Cases

01Scaling production-grade development without frequent manual intervention

02Implementing complex refactorings requiring high quality assurance

03Executing a multi-step feature roadmap with independent components

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tianzecn/myclaudecode subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill