skills.rest
enigmaicon-eng
enigmaicon-engCOMMUNITY
Explore all skills by this author →
GitHub

subagent-driven-development

Execute implementation plans by dispatching fresh subagents per task with two-stage reviews.

1|Updated May 16, 2026
Search Tags:#code review#implementation planning#TDD#agent orchestration#quality gates#subagent#spec compliance
One-click install
npx skills add https://github.com/enigmaicon-eng/AI-Enterprise-OS --skill subagent-driven-development-enigmaicon-eng
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: subagent-driven-development
Source: https://github.com/enigmaicon-eng/AI-Enterprise-OS/tree/main/examples/superpowers/skills/subagent-driven-development
Command: npx skills add https://github.com/enigmaicon-eng/AI-Enterprise-OS --skill subagent-driven-development-enigmaicon-eng

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

It solves the problem of completing a multi-task implementation plan quickly and correctly without context pollution by delegating each task to a fresh subagent and enforcing review gates.

Core Features & Use Cases

  • Fresh subagent per task: Each task runs with isolated context so the implementer doesn’t inherit unrelated session history.
  • Two-stage review per task: Perform a spec-compliance review first, then a code-quality review, and loop until both pass.
  • Continuous plan execution: Run tasks end-to-end from the plan with minimal interruptions, only stopping for genuinely blocking issues.
  • Status-driven recovery: Handle DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED outcomes with the appropriate actions (clarify, provide context, escalate, or split work).
  • Role-based model selection guidance: Use the least powerful model that can handle each role (implementer vs reviewers vs architecture-heavy work).

Quick Start

Tell the AI: “Use subagent-driven-development to execute this implementation plan, running spec then code-quality reviews for each task until everything is approved.”

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 a multi-task implementation plan without context pollution?▼

To execute an implementation plan without context pollution, dispatch a fresh subagent per task with isolated context. This prevents the implementer from inheriting unrelated session history during sequential plan execution.

What is two-stage code review in subagent-driven development?▼

Two-stage code review requires performing a spec-compliance review first, then a code-quality review. The subagent loops both review stages until the implementation passes both quality gates.

How do I handle blocked or incomplete tasks during automated plan execution?▼

Handle blocked or incomplete tasks using status-driven recovery for DONE, DONE_WITH_CONCERNS, NEEDS_CONTEXT, and BLOCKED outcomes. This dictates appropriate actions like clarifying, providing context, escalating, or splitting work.

Can I run implementation tasks end-to-end without manual check-in prompts?▼

Yes, you can run implementation tasks end-to-end with minimal interruptions, only stopping for genuinely blocking issues. This continuous plan execution removes the need for human-style check-in prompts between independent tasks.

When should I not use fresh subagents for implementation planning?▼

Avoid using fresh subagents for implementation planning when tasks are highly interdependent rather than mostly independent, or when your workflow requires frequent manual check-ins rather than continuous same-session execution.

Related Skills

Explore More Software Engineering Skills→
minfengyuanminfengyuan
1

subagent-driven-development

Dispatch isolated subagents to implement tasks with two-stage reviews.

Community
Advanced
jankneumannjankneumann
4

cleanup-feature

Merge approved pull requests and archive OpenSpec proposals to finalize features.

Community
Advanced
TECHKNOWMAD-LABSTECHKNOWMAD-LABS

code-review-engine

Validate code against specifications and analyze quality with severity reports.

Official
Advanced
norrbackanorrbacka

dispatching-parallel-agents

Delegate independent debugging or implementation problems into parallel agent tasks.

Community
Intermediate