What happens if a subagent fails a task or review?

The workflow includes automated re-review loops. If a reviewer finds issues, the implementer subagent is notified to fix them, and the review is repeated until the quality gates are cleared.

How does this differ from the standard 'Executing Plans' skill?

While 'Executing Plans' often runs in parallel sessions, Subagent-Driven Development operates within your current session, providing faster iteration and eliminating the need for context handoffs.

Why are there two separate review stages?

Separating spec compliance from code quality ensures the logic is correct before worrying about style. This prevents wasting time polishing code that doesn't actually meet the requirements.

Does this skill require Test-Driven Development (TDD)?

Yes, subagents are encouraged to follow TDD for each task, ensuring that every implementation is verified by tests before passing through the review gates.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: dkmaker

bydkmaker

•

생산성 및 워크플로

Executes complex implementation plans by dispatching specialized subagents for per-task coding and rigorous two-stage reviews.

Subagent-Driven Development is a high-efficiency workflow for Claude Code designed to execute multi-task implementation plans within a single session. By dispatching a fresh subagent for every task, it prevents context pollution and maintains high focus. The skill enforces a strict quality protocol: every task must pass a specification compliance review followed by a code quality review before moving to the next item. This structured approach minimizes errors, ensures architectural consistency, and automates the tedious aspects of managing large-scale feature deployments.

주요 기능

01Seamless integration with Git worktrees for isolated development

021 GitHub stars

03Automated task tracking and progress management via TodoWrite

04Fresh subagent dispatching per task to maintain clean context

05Interactive Q&A phase to resolve implementation ambiguities upfront

06Mandatory two-stage review process (Spec Compliance and Code Quality)

사용 사례

01Executing multi-step feature plans with independent tasks in a single session

02Refactoring complex modules where each sub-task requires fresh logic and testing

03Scaling AI development while maintaining strict code quality and spec adherence

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dkmaker/my-claude-plugins subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill