subagent-driven-development

Coordinate fresh delegate_task subagents for plan-driven implementation with two-stage review.

Updated May 15, 2026
One-click install
npx skills add https://github.com/cabezno/bmb-encover-agent --skill subagent-driven-development-cabezno
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/cabezno/bmb-encover-agent/tree/main/skills/software-development/subagent-driven-development
Command: npx skills add https://github.com/cabezno/bmb-encover-agent --skill subagent-driven-development-cabezno

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It removes the quality and consistency failures that happen when implementing complex plans in one long, stateful agent run, by enforcing a fresh subagent per task plus a two-stage review before moving on.

Core Features & Use Cases

  • Per-task fresh implementation: Dispatches an implementer subagent for each independent task using the full task context directly in the prompt to avoid accumulated-state confusion.
  • Two-stage review pipeline: Runs a spec compliance reviewer first (verify exactly matches the plan), then a code quality reviewer (tests, conventions, and security-minded checks).
  • Plan-to-todo execution: Reads and parses an implementation plan once, builds a todo list, and tracks task completion with commit-ready expectations.
  • Integration review at the end: Performs a final integration reviewer pass to catch cross-task inconsistencies and confirm tests.

Quick Start

Use the skill to execute an implementation plan by instructing the agent to read your feature plan file once, create a todo list for all extracted tasks, then delegate each task to an implementer subagent followed by spec and quality review until everything is integrated and tests pass.

Frequently Asked Questions about subagent-driven-development

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

FAQPage Schema
How do I implement a complex software plan without losing consistency across tasks?

Implement a complex software plan without losing consistency by dispatching a fresh subagent for each task and enforcing a two-stage review process. This prevents accumulated-state confusion and ensures every task aligns with the spec before advancing.

What is subagent-driven development and how does it work?

Subagent-driven development coordinates fresh delegate_task subagents for each independent task with mandatory two-stage review. It parses an implementation plan into a todo list, runs spec compliance and code quality reviews, and finishes with a final integration review.

How do I automate spec compliance and code quality reviews during feature implementation?

Automate spec compliance and code quality reviews by running a two-stage pipeline after each task implementation. A spec compliance reviewer verifies exact plan matching first, then a code quality reviewer checks tests, conventions, and security before allowing advancement.

Why does long agent execution fail when implementing complex feature plans?

Long agent execution fails due to accumulated-state confusion and quality drift when handling complex feature plans in one run. Using fresh subagents per independent task removes this failure by resetting context and enforcing two-stage reviews before proceeding.

What's the best way to ensure cross-task consistency after plan-driven implementation?

Ensure cross-task consistency after plan-driven implementation by completing a final integration reviewer pass. This catches cross-task inconsistencies, verifies test execution, and confirms all independent tasks integrate correctly before completion.

Do I need to format my implementation plan a specific way for subagent orchestration?

Implementation plans for subagent orchestration must contain explicit tasks that can be parsed once into a todo list. The plan-to-todo execution requires clear task boundaries so each independent task can be delegated directly to an implementer subagent.