subagent-driven-development

Dispatch isolated subagents to execute engineering plans with two-stage reviews.

Updated Mar 16, 2026
One-click install
npx skills add https://github.com/dmulcahey/featureforge --skill subagent-driven-development-dmulcahey
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/dmulcahey/featureforge/tree/main/skills/subagent-driven-development
Command: npx skills add https://github.com/dmulcahey/featureforge --skill subagent-driven-development-dmulcahey

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Execute engineering-approved implementation plans by coordinating a fleet of fresh, isolated subagents for each task and enforcing a disciplined two-stage review process to keep work grounded in a single, plan-driven context.

Core Features & Use Cases

  • Fresh isolated agents per task to prevent cross-task context bleed and enable parallel progress.
  • Two-stage reviews per task: spec compliance review followed by code-quality review to surface gaps early.
  • Plan-driven execution: constructs a task packet from an approved plan/spec pair and treats it as the single source of truth for that task slice.
  • Nested-session guidance and governance: supports multi-task sessions with clear handoffs, guards against drift, and tools for contributor-mode workflows.

Quick Start

Kick off an approved plan and let the system dispatch fresh subagents to work tasks with per-task reviews.

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 engineering plans with isolated subagents to prevent context bleed?

Executing engineering plans with isolated subagents involves dispatching a fresh agent for each task slice derived from an approved plan. This prevents cross-task context bleed by treating the plan as the single source of truth for isolated execution.

What is a two-stage review process for code quality and spec compliance?

A two-stage review process validates engineering work sequentially: first checking spec compliance against the approved plan, then evaluating code quality. This surfaces gaps early and enforces rigorous gatekeeping before protected-branch merges.

How do I coordinate parallel task execution from a single approved implementation plan?

Coordinating parallel task execution from an approved plan requires constructing a task packet for each slice. The system then dispatches dedicated subagents for each task, enabling parallel progress while maintaining plan-driven context.

Can I use subagent orchestration for sessions with mostly independent engineering tasks?

Subagent orchestration is designed for sessions with mostly independent engineering tasks requiring isolated execution contexts. It supports multi-task sessions with clear handoffs and guards against drift through nested-session governance.

When should I not use isolated subagents for task execution?

Isolated subagents should not be used for highly interdependent tasks requiring shared context. The system is built for mostly independent tasks where isolated execution, protected-branch safety gates, and sequential reviews prevent drift.