Why use fresh subagents for every task?

Using a fresh subagent prevents 'context pollution,' ensuring the AI doesn't get confused by previous tasks or irrelevant code changes, leading to higher accuracy.

What happens if a reviewer finds an issue?

The implementation subagent is automatically re-engaged to fix the specific issue. The review cycle then repeats until the reviewer grants full approval.

What is Subagent-Driven Development (SDD)?

SDD is a workflow that breaks down a plan into tasks and assigns each task to a fresh AI subagent. It includes automated reviews for both specification compliance and code quality for every task.

How does the two-stage review process work?

After implementation, a 'Spec Reviewer' ensures the code meets all requirements. Once passed, a 'Code Quality Reviewer' checks for best practices and efficiency. Both must approve before the task is marked complete.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: araaking

byaraaking

•

Productivity & Workflow

Orchestrates specialized subagents to execute complex implementation plans with automated two-stage quality reviews.

Subagent-Driven Development (SDD) is a high-precision workflow skill for Claude Code that automates the execution of multi-task implementation plans. By dispatching fresh subagents for every individual task, SDD prevents context pollution and maintains high architectural standards. The skill enforces a rigorous two-stage review process: first verifying that the implementation strictly matches the technical specifications, and then performing a secondary check for code quality and best practices. This ensures that complex features are built correctly, tested thoroughly, and reviewed exhaustively before the next task begins, all within the same development session.

Key Features

01Automated Quality Loops: Automatically returns tasks to the implementer if reviewers identify gaps or bugs.

02Isolated Subagent Execution: Dispatches fresh agents per task to maintain clean context and focus.

03Session Continuity: Manages task state and progress tracking without requiring context switches or parallel sessions.

04Two-Stage Review Pipeline: Enforces separate specification compliance and code quality checkpoints.

05Interactive Clarification: Enables subagents to ask clarifying questions before work begins to minimize rework.

0618 GitHub stars

Use Cases

01Developing modular systems that require independent implementation and rigorous testing per unit.

02Executing multi-step feature plans where each component has strict architectural requirements.

03Scaling development speed by delegating plan execution to a multi-agent review pipeline.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add araaking/opencode-kuli-pro subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill