subagent-driven-development

Dispatch per-task subagents to execute implementation plans with inline lead review.

2|Updated Mar 3, 2026

Search Tags:#automation #subagents #plan-execution #parallel-dispatch #code-harness #inline-review

Ready to run

npx skills add https://github.com/anortham/razorback --skill subagent-driven-development-anortham

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/anortham/razorback/tree/main/skills/subagent-driven-development
Command: npx skills add https://github.com/anortham/razorback --skill subagent-driven-development-anortham

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Execute an implementation plan by dispatching fresh subagents (sequentially or in parallel) with inline review by the lead. Primary delegated execution path whenever the harness can launch subagents.

Core Features & Use Cases

Fresh subagent per task: Each task is delegated to a new worker to preserve context and avoid cross-task leakage.
Parallel task execution: Independent tasks can dispatch in parallel to speed up delivery.
Inline lead review: The lead performs spec compliance and quality review after each task before proceeding.

Quick Start

Dispatch the plan's tasks to per-task subagents and review inline after each task.

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 using parallel subagents?▼

You can execute an implementation plan by dispatching a fresh subagent for each task, allowing independent tasks to run in parallel. This approach preserves context and avoids cross-task leakage during plan execution.

What is inline lead review in a subagent-driven development workflow?▼

Inline lead review is the process where the lead performs spec compliance and quality verification after each subagent completes its task. This Julie-first orientation ensures changes are reviewed and verified before proceeding to the next task.

Can I use subagent-driven development with different code harnesses like Claude Code or Cursor?▼

Yes, subagent-driven development supports multiple harnesses including Claude Code, Codex, Cursor, OpenCode, Copilot CLI, and Gemini CLI. The workflow dispatches per-task subagents and enforces model routing across these supported environments.

How do I prevent context leakage when delegating multiple implementation tasks to subagents?▼

To prevent context leakage, dispatch a fresh subagent for each implementation task rather than reusing the same worker. This ensures isolated context per task, while the lead performs inline review to maintain spec compliance across the plan.

Does subagent-driven development require specific prompt templates for task execution?▼

Yes, the workflow relies on provided templates such as implementer-prompt.md and fix-prompt.md. These templates guide the per-task subagents and enforce model routing and verification scopes defined in the implementation plan.

What is the best way to speed up delivery when executing an implementation plan with independent tasks?▼

The best way to speed up delivery is dispatching independent tasks in parallel across multiple subagents. This parallel dispatch mechanism allows simultaneous execution while the lead conducts inline review after each task completes.