Is the agent feedback actionable?

Yes, agents use a standardized tagging system (like [CONSTRAINT], [ERROR_STATE], or [REQUIREMENT]) with color-coded UI badges to highlight specific, resolvable issues.

What frameworks are required for this skill?

It is designed for Next.js applications using the App Router, shadcn/ui, and the Vercel AI SDK (ai package).

Can I add my own custom agent personas?

Yes, the skill includes an extensible registry in src/lib/ai/agents/registry.ts where you can define new roles, UI colors, and specialized system prompts.

Does this require multiple LLM calls?

Yes, each agent in the sequence performs an individual API call, passing the accumulated context of the previous agents to the next one to maintain the 'chain of thought'.

How does sequential debate differ from parallel processing?

In this skill, agents run one after another. This allows each subsequent agent to read the previous agents' critiques, preventing redundant feedback and creating deeper, more nuanced pressure-testing.

AI Multi-Agent Debate

Name: AI Multi-Agent Debate
Author: mattwoodco

bymattwoodco

•

협업 도구

Orchestrates a sequential debate between specialized AI personas to pressure-test feature specifications and uncover hidden edge cases.

This skill implements a sophisticated multi-agent orchestration pattern where distinct AI personas—Product Manager, QA Engineer, and Senior Engineer—critique development tasks in a sequential chain. Unlike traditional parallel agents, each participant in this debate sees the full context of previous observations, allowing the QA agent to find gaps in the PM's logic and the Engineer to assess technical feasibility based on functional and quality requirements. This creates an adversarial pressure-testing environment that significantly improves the robustness of feature specs and implementation plans before coding begins.

주요 기능

01Sequential agent orchestration for deep, context-aware debate

02Extensible agent registry for adding custom domain-specific roles

03Real-time Server-Sent Events (SSE) for live streaming agent responses

043 GitHub stars

05Pre-configured specialized personas (PM, QA, and Senior Engineer)

06Actionable tagging system (e.g., [EDGE_CASE], [CONSTRAINT]) for easy triaging

사용 사례

01Simulating a cross-functional team review for solo developers or small teams

02Generating comprehensive acceptance criteria and implementation strategies

03Pressure-testing new feature specifications for hidden technical debt or edge cases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add mattwoodco/skills ai-multi-agent

For use in Claude.ai and ChatGPT

Download Skill