What happens if a reviewer finds an issue?

The workflow triggers an automatic fix-and-review cycle. The implementation subagent addresses the feedback, and the reviewer re-evaluates the code until it passes all checks.

Do I need a pre-existing plan to use this skill?

Yes, this skill is designed to execute an implementation plan. It works best when paired with the 'writing-plans' skill to generate the roadmap first.

Is context shared between subagents?

The controller subagent provides necessary context to each implementer, but each task subagent starts with a 'fresh' state to prevent errors caused by irrelevant previous task history.

How does this differ from standard plan execution?

Unlike standard execution, this skill stays in the same session while using fresh subagents for each task, adding a mandatory two-stage review loop (spec and quality) for significantly higher reliability.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: dazeb

bydazeb

•

생산성 및 워크플로

Executes complex implementation plans by dispatching specialized subagents for each task with rigorous two-stage review loops.

Subagent-Driven Development is a specialized workflow designed to execute detailed implementation plans within a single Claude session. By breaking plans into independent tasks and dispatching a fresh subagent for each, the skill prevents context pollution and maintains high focus. Every task undergoes a mandatory two-stage review: first, a specification compliance check to ensure all requirements are met, and second, a code quality review to maintain architectural standards. This approach combines the speed of automated iteration with the reliability of multi-agent validation, making it ideal for high-stakes feature development.

주요 기능

014 GitHub stars

02Two-stage verification process: Spec compliance followed by code quality review.

03Automated feedback loops where implementers fix issues identified by reviewers.

04Fresh subagent context for every task to eliminate logic bleed and confusion.

05Integrated task management and progress tracking using TodoWrite.

06Support for TDD (Test-Driven Development) within each subagent task.

사용 사례

01Executing a multi-task feature plan while maintaining a single session history.

02Building complex software components where spec-adherence is critical for success.

03Implementing modular architectural changes that require high precision and review.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add dazeb/claudeplaybook subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill