subagent-driven-development

Dispatch isolated subagents per task and validate through sequential spec and code reviews.

6|Updated Mar 21, 2026
One-click install
npx skills add https://github.com/caitengwei/my-skills --skill subagent-driven-development-caitengwei
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/caitengwei/my-skills/tree/main/superpowers/skills/subagent-driven-development
Command: npx skills add https://github.com/caitengwei/my-skills --skill subagent-driven-development-caitengwei

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Subagent-driven development solves the problem of implementing complex plans reliably by delegating each task to a fresh, focused agent context and enforcing checkpoints so you catch spec mistakes and code-quality issues early.

Core Features & Use Cases

  • Fresh subagent per task: Prevents context pollution by ensuring each implementer starts with only the instructions and context it needs for that specific task.
  • Two-stage review loop: After each task, runs a spec-compliance review first, then a code-quality review, with fix-and-re-review until approved.
  • Status-based control: Uses DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED to decide whether to proceed, request missing info, escalate, or refine task scope.
  • When to use: Ideal for executing well-defined implementation plans where tasks are mostly independent and you want fast iteration without continuous manual handoffs.

Quick Start

Use subagent-driven-development to execute an implementation plan by dispatching an implementer subagent for each task, running spec review then code quality review after each task, and iterating on fixes until both reviews approve.

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 multi-step implementation plans without context pollution?

Dispatch isolated subagents per task to execute implementation plans, ensuring each implementer starts with fresh context containing only task-specific instructions to prevent context pollution and maintain focus.

What is the two-stage code review process for subagent-driven development?

The two-stage code review process runs a spec-compliance review first, then a code-quality review, iterating on fix-and-re-review cycles until both reviews approve the task outcome.

How do I handle blocked tasks or missing context in workflow orchestration?

Use status-based control with DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED statuses to decide whether to proceed, request missing information, escalate, or refine task scope during workflow orchestration.

When should I use subagent-driven development for test-driven development workflows?

Use subagent-driven development when executing well-defined implementation plans where tasks are mostly independent, require spec-accuracy enforcement, and benefit from cost- and speed-aware model selection.

Does subagent-driven development require dependencies or external components?

No, subagent-driven development operates with zero dependencies and no external components, requiring only per-task fresh context and iterative review handling to function independently.