subagent-driven-development

Execute implementation plans by delegating tasks to subagents with two-stage review.

Updated Apr 9, 2026
One-click install
npx skills add https://github.com/MarbleSodas/Mavis --skill subagent-driven-development-marblesodas
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/MarbleSodas/Mavis/tree/main/skills/software-development/subagent-driven-development
Command: npx skills add https://github.com/MarbleSodas/Mavis --skill subagent-driven-development-marblesodas

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It solves the problem of turning an implementation plan into correct, maintainable code without relying on a single long-running agent session that can drift or accumulate errors.

Core Features & Use Cases

  • Two-stage review per task: runs a spec compliance check first, then a code quality review second to prevent scope creep and catch defects early.
  • Fresh implementer per task: dispatches a clean subagent for each task so context pollution is minimized and each change is tightly scoped.
  • Task-level verification and commit discipline: encourages failing-test-first (TDD), running tests, and committing at task completion boundaries.

Use case example: You have an implementation plan for a small auth feature (e.g., user model, password hashing, login endpoint). Use this skill to execute each task with an implementer subagent, then ensure the output matches the plan spec and passes quality checks before moving on.

Quick Start

Ask the AI to run subagent-driven-development on your feature implementation plan and implement each task with TDD plus spec compliance and code quality review gates.

Frequently Asked Questions about subagent-driven-development

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

FAQPage Schema
How do I execute an implementation plan without agent context drift?

Subagent-driven development executes implementation plans by delegating each task to a fresh implementer subagent, minimizing context pollution and preventing error accumulation in long-running agent sessions.

What is the best way to automate spec compliance and code quality reviews?

Spec compliance and code quality reviews are automated through a two-stage review gate, enforcing spec checks first and quality checks second, then re-running both reviews after any fixes are applied.

Can I use TDD with subagent delegation for feature implementation?

TDD is supported during subagent delegation by encouraging a failing-test-first approach, running tests, and committing at task completion boundaries to maintain strict verification discipline.

How do I prevent scope creep when delegating implementation tasks to subagents?

Scope creep is prevented by parsing all tasks up front, using delegate_task with full per-task context, and enforcing a strict review ordering where spec compliance is verified before code quality.

Does subagent-driven development work for independent task batches from writing-plans?

Subagent-driven development is suitable for independent task batches derived from writing-plans or user requirements where high fidelity and fast iteration matter.

Why does my code review fail to catch defects early in implementation workflows?

Defects are caught early by applying a two-stage review for spec compliance and code quality per task, ensuring each implementation step is tightly scoped and approved before moving on.