subagent-driven-development

Dispatch subagents for isolated plan tasks with two-stage reviews.

2|1|Updated Dec 9, 2025
One-click install
npx skills add https://github.com/ninjasitm/ai-assisted-dev-toolkit --skill subagent-driven-development-ninjasitm
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/ninjasitm/ai-assisted-dev-toolkit/tree/main/src/monorepo/.agents/skills/subagent-driven-development
Command: npx skills add https://github.com/ninjasitm/ai-assisted-dev-toolkit --skill subagent-driven-development-ninjasitm

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps teams execute complex implementation plans by dispatching fresh subagents for independent tasks within the same session, while maintaining high quality through a two-stage review process.

Core Features & Use Cases

  • Fresh subagent per task to avoid context pollution and enable parallel progress
  • Two-stage review after each task: spec compliance, then code quality
  • Per-task orchestration within a single session with automatic Todo tracking
  • Flexible use cases for planning and executing multi-step features

Quick Start

Initiate the workflow by dispatching the first implementer subagent for Task 1 and proceed through the spec compliance and code quality reviews.

Frequently Asked Questions about subagent-driven-development

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I use subagents to execute multi-step development plans?

You can dispatch fresh subagents for independent tasks within a single session to execute multi-step development plans. This approach enables rapid, iterative execution with swift feedback by giving each task an isolated context.

What is a two-stage review process for code implementation?

A two-stage review process checks task implementation in two phases: spec compliance and code quality. It operates through a guarded review loop using spec-reviewer and code-quality-reviewer prompts after each subagent completes its assigned task.

How do I prevent context pollution when running parallel code generation tasks?

To prevent context pollution during parallel code generation, dispatch a fresh subagent per task. This isolates the execution context, allowing independent tasks to progress without interfering with each other's working memory.

Does subagent-driven development work for dependent tasks in the same session?

Subagent-driven development is designed for in-session plans where tasks are mostly independent. It may not be suitable for dependent tasks, as it relies on dispatching isolated subagents for rapid, iterative execution.

What is the best way to orchestrate feature implementation with isolated subagents?

The best way to orchestrate isolated subagents is to dispatch focused subagents for each task and proceed through spec compliance and code quality reviews. This workflow includes automatic Todo tracking to manage per-task orchestration.

When should I use a fresh subagent per task instead of a single agent?

Use a fresh subagent per task when you need to execute complex implementation plans with independent tasks requiring rapid, iterative execution and swift feedback. A single agent risks context pollution across multiple plan tasks.