subagent-driven-development

Dispatch fresh subagents for each task in an implementation plan.

4|2|Updated May 4, 2026
One-click install
npx skills add https://github.com/zpyoung/quirk --skill subagent-driven-development-zpyoung
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/zpyoung/quirk/tree/main/skills/subagent-driven-development
Command: npx skills add https://github.com/zpyoung/quirk --skill subagent-driven-development-zpyoung

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires Task, Task, Task, mcp__pal__clink, Task, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill simplifies the execution of complex implementation plans with independent tasks, ensuring focused, high-quality iterations in the same session.

Core Features & Use Cases

  • Task Distribution: Dispatches fresh subagents per task, maintaining isolated context.
  • Three-Pass Review: Ensures spec compliance, code quality, and Codex adversarial gap-finding after each task.
  • Mode Selection: Supports sequential, in-place parallel, worktree parallel, and team execution modes.
  • Use Case: Ideal for parallel agent dispatch and code review in a single session, reducing context switching and increasing iteration speed.

Quick Start

To initiate subagent-driven development, specify the task and the plan using the /quirk:subagent-driven-development command.

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 with independent tasks in parallel?

You can dispatch fresh subagents for each task with independent specifications, ensuring focused execution and comprehensive reviews within a single session. This Skill supports sequential, in-place parallel, worktree parallel, and team execution modes.

What is subagent-driven development for parallel code review and task execution?

Subagent-driven development dispatches isolated subagents per task to maintain clean context, followed by a three-pass review checking spec compliance, code quality, and adversarial gap-finding after each task completes.

Which agents can I use for parallel subagent dispatch and code review?

You need Claude or Pi agents for execution and review when dispatching subagents. The Skill relies on the Task tool and mcp__pal__clink dependency to coordinate these independent subagent workers.

Can I run parallel task execution using git worktrees with subagents?

Yes, you can select worktree parallel execution mode to dispatch isolated subagents across separate worktrees. This allows concurrent task execution while maintaining focused context for each independent specification.

What is the best way to review code quality across multiple parallel agent tasks?

A three-pass review system ensures spec compliance, code quality, and Codex adversarial gap-finding after each dispatched subagent finishes its task. This comprehensive review process reduces context switching and increases iteration speed.

Why does my implementation plan fail when tasks share dependencies instead of being independent?

This Skill applies to tasks with independent specifications only. Dispatching fresh subagents requires isolated context for each task, so shared dependencies or overlapping execution states will break the focused review and parallel processing workflow.