subagent-driven-development

Dispatch a fresh subagent per task with two-stage spec and code reviews.

Updated Jan 19, 2026
One-click install
npx skills add https://github.com/kgadek/kgadek-claude-plugin --skill subagent-driven-development-kgadek
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/kgadek/kgadek-claude-plugin/tree/main/ai-spec/2026-01-21-sdd-inspirations/references/obra-superpowers/skills/subagent-driven-development
Command: npx skills add https://github.com/kgadek/kgadek-claude-plugin --skill subagent-driven-development-kgadek

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Coordinating multi-task implementations within a single session can lead to context leakage, inconsistent progress, and slower iterations. Subagent-driven development solves this by dispatching a fresh subagent per task and applying a two-stage review to ensure spec compliance before code quality checks.

Core Features & Use Cases

  • Fresh subagent per task to isolate context and prevent cross-task pollution
  • Two-stage review: spec compliance first, then code quality
  • Deterministic per-task workflow with clear handoffs and iterative fixes
  • Use Case: executes complex plans with mostly independent tasks requiring fast turnaround within the same session

Quick Start

Dispatch an implementer subagent for Task N with the full task text and context to begin execution.

Frequently Asked Questions about subagent-driven-development

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

FAQPage Schema
How do I prevent context leakage when automating multi-task workflows in a single session?

Subagent-driven development prevents context leakage by dispatching a fresh subagent per task, isolating execution to avoid cross-task pollution while maintaining rapid iteration within the same session.

What is the two-stage review process for task implementation?

The two-stage review process enforces spec compliance verification first, followed by a code quality review, creating deterministic per-task workflows with iterative rework loops as needed.

Can I use subagent task dispatch for dependent tasks within the same session?

Subagent task dispatch is best for mostly independent tasks requiring fast turnaround. Coordinating dependent tasks is not ideal due to the strict fresh subagent per task isolation model.

What's the best way to automate spec compliance and code review for rapid iteration?

Automating spec compliance and code review is best handled by dispatching a fresh implementer subagent per task with a strict two-stage review workflow to enforce deterministic quality checks.

Why does my multi-task plan execution suffer from inconsistent progress?

Inconsistent progress during multi-task plan execution often stems from context pollution. Using a fresh subagent per task with strict two-stage reviews isolates context and ensures deterministic handoffs.

Do I need external dependencies to run subagent-driven task management?

No external dependencies are required. Subagent-driven task management operates independently to dispatch implementation subagents and execute spec and code quality reviews with rework loops.