Can this skill handle complex dependencies between tasks?

It is best suited for plans with mostly independent tasks. If tasks are tightly coupled or require frequent plan revisions, manual execution or brainstorming is recommended.

How does Subagent-Driven Development differ from normal plan execution?

Unlike manual execution, this skill uses a 'fresh' subagent for every single task to avoid context pollution and forces a mandatory code review step between every task to ensure bugs are caught early.

Does this skill require other plugins or skills?

Yes, it integrates with writing-plans to start, requesting-code-review for quality gates, and finishing-a-development-branch to finalize work.

Is this workflow more expensive in terms of token usage?

While it uses more subagent calls, it significantly reduces costs associated with long-term debugging by enforcing quality and testing at every stage of development.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: NickCrew

byNickCrew

•

生产力与工作流

Orchestrates specialized subagents to execute complex implementation plans with automated code reviews between tasks.

Subagent-Driven Development is an advanced workflow for Claude Code that executes structured plans by dispatching a fresh, isolated subagent for every individual task. This approach prevents context pollution and ensures high code quality by automatically triggering independent code reviews between every step. It streamlines the development lifecycle within a single session, combining Test-Driven Development (TDD) principles with systematic quality gates to deliver production-ready features rapidly and reliably without requiring constant human-in-the-loop intervention.

主要功能

01Same-Session Execution: Maintains progress and state without the need for manual context switching.

02Integrated Quality Gates: Automatically triggers code reviews after every task implementation.

03TDD-First Approach: Ensures every task is verified with tests before completion.

04Fresh Subagent Orchestration: Dispatches isolated agents for each task to prevent context drift.

05Automated Issue Resolution: Dispatches follow-up subagents to fix review findings before moving forward.

067 GitHub stars

使用场景

01Maintaining high code quality across large-scale refactors using automated review checkpoints.

02Scaling development speed by allowing the AI to iterate through plan tasks with minimal human intervention.

03Executing complex multi-task feature implementations within a single development session.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nickcrew/claude-cortex subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill