adversarial-verifier

Decompose claims and independently verify code, analyses, and architecture.

1|Updated Jan 11, 2026

Search Tags:#edge-cases #verify #code-analysis #adversarial #architecture-review #independent-verification #claims-decomposition

Ready to run

npx skills add https://github.com/dazuck/claude-code-skills --skill adversarial-verifier

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: adversarial-verifier
Source: https://github.com/dazuck/claude-code-skills/tree/main/skills/adversarial-verifier
Command: npx skills add https://github.com/dazuck/claude-code-skills --skill adversarial-verifier

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill acts as a deeply skeptical verification agent that finds hidden, pernicious errors in code, analysis, and architecture—questions every assumption and refuses to take surface correctness at face value.

Core Features & Use Cases

Independent verifier that challenges every claim and traces reasoning for hidden flaws.
Deep drilling across code, data analyses, and architectural designs to surface edge cases and silent errors.
Use cases include code reviews, model analyses, and system designs where incorrect assumptions may propagate.

Quick Start

Activate adversarial-verifier and instruct it to decompose claims, generate verification questions, and perform independent investigations before synthesizing findings.

Frequently Asked Questions about adversarial-verifier

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

FAQPage Schema

How do I verify code analysis claims for hidden errors?▼

Verifying code analysis claims for hidden errors requires an adversarial approach that decomposes claims, generates verification questions, and performs independent investigations to synthesize accurate findings. This skill implements the Chain-of-Verification protocol to challenge surface correctness and expose pernicious flaws.

What is adversarial verification in software engineering?▼

Adversarial verification in software engineering is a methodical process of challenging every assumption in code, data analyses, and architecture to uncover silent errors. It traces reasoning deeply and refuses to take surface correctness at face value, ensuring that incorrect assumptions do not propagate.

How do I perform an independent architecture review to find edge cases?▼

Performing an independent architecture review to find edge cases involves deep drilling across system designs to surface hidden flaws. This skill acts as a skeptical verifier that decomposes architectural claims, questions every assumption, and independently investigates them before synthesizing a final assessment.

Can I use this to catch silent errors in data analyses?▼

Yes, you can use this to catch silent errors in data analyses. It applies deep drilling to surface edge cases and incorrect assumptions that pass surface inspection, ensuring that analytical claims are independently verified rather than blindly trusted.

What's the best way to review code for pernicious flaws that pass surface inspection?▼

The best way to review code for pernicious flaws is applying the Chain-of-Verification protocol: decompose claims, generate verification questions, conduct independent investigations, and synthesize findings. This ruthlessly exposes hidden errors that standard code reviews might miss.

When should I not use an adversarial approach for code verification?▼

You should not use an adversarial approach for code verification when you need a quick surface-level syntax check or standard functional review. This method is designed for deep drilling into complex code, analyses, and system designs where hidden assumptions and silent errors are highly probable.