How does this differ from standard plan execution?

Unlike standard execution, this skill uses fresh subagents for every task to prevent context drift and mandates a two-stage review process (spec then quality) to ensure higher reliability.

What happens if a subagent finds an issue during review?

The workflow triggers a fix-and-review loop where the implementer subagent addresses the feedback and the reviewer re-evaluates the code until all gates are passed.

Does this skill require an existing plan?

Yes, it is designed to execute structured implementation plans where tasks are relatively independent and can be isolated.

Why is the review process split into two stages?

Separating spec compliance from code quality ensures the subagent first builds the right thing before the workflow focuses on the technical excellence of the implementation.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: yang0

byyang0

0•

Productivity & Workflow

Executes complex implementation plans by dispatching isolated subagents with a rigorous two-stage review process.

Subagent-Driven Development is a sophisticated workflow skill designed to execute multi-task implementation plans within a single Claude session. By dispatching a fresh subagent for every task, it prevents context pollution and AI drift, ensuring each component is built with maximum focus. The skill enforces a mandatory two-stage review process for every task: first verifying strict compliance with the project specifications, followed by a dedicated code quality assessment. This methodology accelerates development cycles while maintaining high-quality standards through automated checkpoints and integrated Test-Driven Development (TDD).

Key Features

01Full integration with Git worktrees and TDD principles

02Two-stage review pipeline: Spec Compliance first, Code Quality second

03Pre-implementation question phase to resolve ambiguities early

040 GitHub stars

05Fresh subagent dispatch per task to eliminate context pollution

06Automated task extraction and checklist management from implementation plans

Use Cases

01Automating the implementation and review cycle for large-scale refactors

02Maintaining high code standards in fast-paced AI-assisted development sessions

03Executing complex multi-feature plans where tasks are modular and independent

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yang0/superpower subagent-driven-development

For use in Claude.ai and ChatGPT

Key Features

01Full integration with Git worktrees and TDD principles

02Two-stage review pipeline: Spec Compliance first, Code Quality second

03Pre-implementation question phase to resolve ambiguities early

040 GitHub stars

05Fresh subagent dispatch per task to eliminate context pollution

06Automated task extraction and checklist management from implementation plans

Use Cases

01Automating the implementation and review cycle for large-scale refactors

02Maintaining high code standards in fast-paced AI-assisted development sessions

03Executing complex multi-feature plans where tasks are modular and independent

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yang0/superpower subagent-driven-development

For use in Claude.ai and ChatGPT