adversarial-claims-reviewer

Inventory equations and claims, assign IDs, and classify verdicts.

15|3|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/LazyIsEfficient/agentic-os --skill adversarial-claims-reviewer
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: adversarial-claims-reviewer
Source: https://github.com/LazyIsEfficient/agentic-os/tree/main/.claude/skills/adversarial-claims-reviewer
Command: npx skills add https://github.com/LazyIsEfficient/agentic-os --skill adversarial-claims-reviewer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sympy, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Adversarial-claims-reviewer addresses the challenge of validating formal or quantitative claims in documents by forcing a rigorous inventory and verdict process, ensuring that each assertion is checked and properly classified.

Core Features & Use Cases

  • INVENTORY every displayed equation, quantitative claim, and named theorem-use, assigning stable IDs and building a traceable audit trail.
  • RESTATE each identified claim as a precise, self-contained proposition with explicit symbol definitions and domains.
  • VERIFY claims using deterministic symbolic computation, fixed-point spot-checks, and known identities, complemented by regime sanity checks.
  • CLASSIFY every claim as VERIFIED, REFUTED, UNVERIFIABLE, or VACUOUS, with per-claim justifications and executable verification scripts.
  • REGIME SANITY to compare formulas against known regimes (e.g., limits, trivial solutions) to ensure correct behavior.
  • SELF-CONSISTENCY SWEEP to detect mismatches between body assertions and appendices, and to ensure consistent symbol usage.
  • REPORT generation against a structured template, including a concise most damaging finding and a detailed claims table.

Quick Start

Run the adversarial-claims-reviewer on a document to produce a complete inventory, per-claim verdicts, and a structured report.

Frequently Asked Questions about adversarial-claims-reviewer

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

FAQPage Schema
How do I verify equations and quantitative claims in a technical document?

You can audit formal claims by inventorying every displayed equation and quantitative assertion, assigning stable IDs, and applying deterministic symbolic computation to classify each claim as VERIFIED, REFUTED, UNVERIFIABLE, or VACUOUS.

What is the best way to check mathematical papers for self-consistency and theorem misuse?

The best way is to perform a self-consistency sweep that restates claims as precise propositions, compares formulas against known regimes, and detects mismatches between body assertions and appendices for consistent symbol usage.

Can I use sympy to automate the verification of physics formulas and benchmarking results?

Yes, you can use sympy to automate the verification of physics formulas and benchmarking results through deterministic symbolic computation, fixed-point spot-checks, and known identities, complemented by regime sanity checks.

How do I generate a structured audit report for formal claims with per-claim verdicts?

You generate a structured audit report by classifying each formal claim with per-claim justifications, executing verification scripts, and populating a final exportable template that includes a concise most damaging finding and a detailed claims table.

Does adversarial document review work for checking limits and trivial solutions in technical papers?

Yes, adversarial document review applies regime sanity checks that compare formulas against known regimes, such as limits and trivial solutions, to ensure correct behavior and validate assertions in technical papers.

What are the limitations of using symbolic computation for unverifiable mathematical claims?

Limitations occur when claims lack explicit symbol definitions or domains, rendering them UNVERIFIABLE or VACUOUS despite deterministic protocols, which requires manual restatement and domain specification to proceed.