debate

Orchestrate multi-model expert challenges against Claude's position and produce ADR-ready decision records.

2|Updated Jan 7, 2026

Search Tags:#adr #audit #multi-model #consensus #debate #adversarial #decision-support

Ready to run

npx skills add https://github.com/nalyk/nalyk-skills-demo --skill debate-nalyk

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: debate
Source: https://github.com/nalyk/nalyk-skills-demo/tree/main/plugins/debate
Command: npx skills add https://github.com/nalyk/nalyk-skills-demo --skill debate-nalyk

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Multi-model adversarial debate for high-stakes decisions, surfacing blind spots and diverse perspectives to improve outcomes.

Core Features & Use Cases

Orchestrates three distinct expert challengers (Architect, Operator, Adversary) to critique Claude's position.
Produces auditable outputs including consensus checks, evolved positions, and ADR-ready tradeoffs.
Provides prefight checks, parallel challenge phase, confrontation rounds, and a formal conclusion (consensus or tradeoff).

Quick Start

Initiate a debate by providing a topic to trigger a multi-model critique and produce an auditable decision record.

Frequently Asked Questions about debate

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

FAQPage Schema

How does multi-model debate improve architecture and security decisions?▼

Multi-model debate improves architecture and security decisions by orchestrating parallel critiques from distinct expert challengers against an initial position. This adversarial process surfaces blind spots and diverse perspectives, evolving the original stance into a robust, consensus-driven outcome.

How do I generate an auditable ADR for a product strategy decision?▼

To generate an auditable ADR for product strategy, initiate a debate by providing your topic. The process runs prefight checks and parallel critiques, concluding with a formal consensus or tradeoff output that is directly formatted as an Architecture Decision Record.

Can I configure the confrontation rounds and challenger behavior locally?▼

Yes, you can configure the debate behavior locally. The Skill supports a local debate.local.md file, allowing you to adjust parameters for prefight checks, parallel challenge phases, and optional confrontation rounds to suit your decision-support workflow.

What is the difference between consensus checks and tradeoff outputs in adversarial decision support?▼

Consensus checks determine if challenger critiques align with the original position to reach agreement, while tradeoff outputs document unresolved differences as formal concessions. Both provide auditable ADR-ready records for high-stakes architecture and security decisions.

When should I use multi-model adversarial critique instead of a single expert analysis?▼

Use multi-model adversarial critique for high-stakes architecture, product strategy, and security decisions where diverse perspectives reduce blind spots. Rely on single expert analysis for straightforward tasks lacking the complexity to warrant parallel challenger confrontation rounds.