What happens if a subagent fails a task or review?

The workflow includes an automatic feedback loop where a fix subagent is dispatched to address specific issues identified during the code review phase before moving to the next task.

How does this differ from standard plan execution?

Unlike basic execution, this skill dispatches a fresh subagent for every task within the same session, preventing context pollution and including mandatory code reviews between steps.

Can I use this for large features?

This is the recommended skill for large features, as it breaks the work into manageable, reviewed chunks, reducing the risk of compounding errors.

Do I need a plan before using this skill?

Yes, this skill is designed to execute a pre-existing plan, typically generated by the 'writing-plans' skill, which outlines the individual tasks to be completed.

Does it support Test-Driven Development (TDD)?

Yes, subagents are explicitly instructed to write and verify tests as part of their implementation task, ensuring high code reliability.

Subagent-Driven Development

Name: Subagent-Driven Development
Author: dsifry

bydsifry

•

Productivity & Workflow

Automates the execution of multi-step implementation plans by dispatching fresh subagents for each task followed by iterative code reviews.

About

Subagent-Driven Development is a sophisticated workflow skill for Claude Code that orchestrates the execution of complex coding plans through isolated sub-tasks. By assigning a fresh subagent to each plan item and enforcing automated code reviews between steps, it prevents context pollution and ensures high-quality output with rigorous quality gates. This approach combines the speed of automated iteration with the reliability of continuous integration, making it ideal for large-scale feature development where maintainability and architectural integrity are paramount.

Key Features

Comprehensive final validation to ensure all plan requirements are met
Closed-loop feedback for immediate resolution of review findings
Seamless integration with Test-Driven Development (TDD) principles
Isolated subagent execution per task to prevent context drift and pollution
Mandatory automated code reviews after every implemented step
29 GitHub stars

Use Cases

Executing complex feature plans with multiple independent modules or tasks
Automating high-quality code migrations and large-scale refactors
Maintaining strict architectural standards through automated quality gates during development

About

Key Features

Comprehensive final validation to ensure all plan requirements are met
Closed-loop feedback for immediate resolution of review findings
Seamless integration with Test-Driven Development (TDD) principles
Isolated subagent execution per task to prevent context drift and pollution
Mandatory automated code reviews after every implemented step
29 GitHub stars

Use Cases

Executing complex feature plans with multiple independent modules or tasks
Automating high-quality code migrations and large-scale refactors
Maintaining strict architectural standards through automated quality gates during development