What is the benefit of using fresh subagents per task?

Using fresh subagents ensures that the AI's context window remains clean and focused solely on the current task, significantly reducing hallucinations and 'context drift' that occurs in long-running sessions.

Do I need other skills to use Subagent-Driven Development?

Yes, it is designed to integrate with other h-superpowers skills like 'writing-plans' for the initial roadmap and 'finishing-a-development-branch' for final merging.

How does this differ from standard plan execution?

Unlike standard execution, this skill stays within the current session for faster iteration while using fresh subagents for each task to prevent context pollution and adding mandatory automated review checkpoints.

What happens during the two-stage review process?

First, a 'Spec Reviewer' performs an adversarial check to ensure the code matches the requirements line-by-line. Only after passing is a 'Code Quality Reviewer' dispatched to check for clean code and maintainability.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: ehartye

byehartye

0•

Productivity & Workflow

Executes complex implementation plans by dispatching specialized subagents for coding and multi-stage reviews to ensure high-quality, spec-compliant output.

Subagent-Driven Development is a sophisticated execution framework for Claude Code that breaks down implementation plans into independent tasks handled by ephemeral subagents. By utilizing a dedicated implementer followed by mandatory spec compliance and code quality reviews for every task, it prevents context pollution and ensures rigorous adherence to requirements. This skill is ideal for developers who want to maintain high velocity in a single session while enforcing strict quality gates and Test-Driven Development (TDD) principles without the overhead of manual context switching.

Key Features

01Task-specific subagent dispatching to prevent context drift and pollution

02Seamless workflow orchestration from initial plan extraction to final branch merge

030 GitHub stars

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

05Mandatory two-stage review process focusing on spec compliance then code quality

06Automated task tracking and progress management via TodoWrite integration

Use Cases

01Enforcing strict adherence to technical specifications through adversarial review subagents

02Executing complex multi-task feature plans within a single Claude session

03Scaling development velocity by automating the verification and quality control loops

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ehartye/hartye-superpowers subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill