How does Subagent-Driven Development prevent context pollution?

It dispatches a completely fresh subagent for every task in your plan. This ensures the AI doesn't get confused by previous task details or accumulated conversational noise, leading to cleaner and more accurate code.

Why is code review required between every task?

Reviewing after each task acts as a quality gate. It allows the system to catch logic errors or architectural drifts immediately, preventing small mistakes from compounding into larger issues later in the plan.

Can I use this skill without a pre-written plan?

No, this skill specifically requires a plan file created by the writing-plans skill. It uses that plan as a roadmap to track progress and define subagent tasks.

What is the TDD Compliance Certification?

It is a mandatory report generated by the subagent for every new function. It proves that the function was built following Test-Driven Development principles and has passed its associated tests.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: bacchus-labs

bybacchus-labs

•

生産性とワークフロー

Executes implementation plans by dispatching isolated subagents for individual tasks with integrated code reviews and quality gates.

Subagent-Driven Development is a high-velocity workflow designed for Claude Code that breaks down complex plans into modular, independent tasks. By dispatching a fresh subagent for every task, the skill prevents context pollution and ensures that each feature is implemented with specialized focus. The process enforces strict quality control by requiring Test-Driven Development (TDD) compliance and mandatory automated code reviews between every task. This approach allows for rapid, continuous progress within a single session while catching architectural issues and bugs early in the development cycle.

主な機能

01Mandatory TDD Compliance Certification for all implemented functions

02Final architectural validation before completing the development branch

03Isolated subagent dispatching to maintain clean context for every task

04Dynamic feedback loops for immediate correction of critical issues

053 GitHub stars

06Automated code review checkpoints between tasks to ensure quality

ユースケース

01Implementing complex multi-step features that require high architectural integrity

02Executing detailed development plans without manual intervention between tasks

03Maintaining strict TDD and code review standards during rapid AI-assisted prototyping

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bacchus-labs/wrangler subagent-driven-development

For use in Claude.ai and ChatGPT

Download Skill