adversarial-thinking

Generate three divergent candidates and select a winner through isolated critique cycles.

Updated Apr 8, 2026
One-click install
npx skills add https://github.com/rd162/skills --skill adversarial-thinking-rd162
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: adversarial-thinking
Source: https://github.com/rd162/skills/tree/main/adversarial-thinking
Command: npx skills add https://github.com/rd162/skills --skill adversarial-thinking-rd162

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Adversarial-thinking produces rigorously stress-tested solutions by generating divergent approaches and evaluating them through isolated dialogues between attacker and defender agents, then selecting the strongest via pairwise comparison. It helps ensure robust answers for high-stakes questions.

Core Features & Use Cases

  • Divergent candidate generation and isolated critique to surface multiple high-quality solutions.
  • Pairwise comparison and winner/runner-up selection to minimize risk of flawed decisions.
  • Use cases include architecture decisions, complex strategy planning, and risk assessment in high-stakes domains.

Quick Start

Provide three divergent approaches and run the adversarial-thinking pipeline to identify the best option

Frequently Asked Questions about adversarial-thinking

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

FAQPage Schema
How do I stress-test architecture decisions before committing to a solution?

Stress-test architecture decisions by generating three divergent candidates and evaluating them through isolated critique and revision cycles. This pipeline selects a winner and runner-up via pairwise comparison to ensure robust outcomes.

What is adversarial thinking in multi-agent critique and how does it work?

Adversarial thinking uses multi-agent critique where isolated attacker and defender agents evaluate divergent approaches. It enforces phase-based sub-agent isolation and Phase 2.5 checks to guide evaluation using enriched requirements with anti-requirements.

How do I use divergent candidate generation to minimize risk in complex strategy planning?

Use divergent candidate generation to produce multiple approaches, then run them through isolated critique cycles. Pairwise comparison of the revised candidates identifies the strongest solution, minimizing the risk of flawed strategic decisions.

Can I use multi-agent critique for risk assessment in high-stakes domains?

Yes, multi-agent critique supports risk assessment in high-stakes domains by isolating attacker and defender dialogues to evaluate candidates. It outputs a rigorously stress-tested winner and runner-up for complex strategy planning.

What are the limitations of using isolated critique cycles for solution verification?

Isolated critique cycles require enriched requirements with anti-requirements to guide evaluation effectively. Without providing three divergent approaches initially, the pairwise comparison cannot adequately select a winner to minimize decision risk.