Can this skill be used for general software testing?

While it is optimized for the 'finance-driver' and real estate underwriting, its framework for checking reasonableness and edge cases is highly effective for any data-intensive software project.

Do I need external tools for the visual validation features?

Yes, to capture screenshots and visual evidence, the Playwright MCP server must be installed and configured in your Claude environment.

Where are the validation results stored?

All results, status tables, and links to captured evidence are consolidated into a 'validation.md' file located within your project's directory for easy tracking and export.

How does this skill prevent AI hallucinations in financial formulas?

The skill employs a systematic 'AI-specific check' and a specialized PAL (Programmatic Abstraction Layer) code review to verify logic chains, cite sources, and cross-reference formulas against authoritative benchmarks.

What is the 'Iron Law' in the validation process?

The Iron Law dictates that a pilot should never trust a single instrument; similarly, this skill requires checking known answers, reasonableness, edge cases, and AI-specific risks for every implementation.

Financial Model & Code Validator

Name: Financial Model & Code Validator
Author: peterjohn1298

bypeterjohn1298

0•

セキュリティとテスト

Ensures financial models and underwriting tools are accurate, reasonable, and defensible through systematic cross-checking and automated testing.

The Validate skill transforms Claude into a rigorous 'Cognition Mate' for developers and analysts, providing a structured framework to verify complex financial underwriting tools. It enforces a professional 'Iron Law' of validation, requiring four independent checks: matching known benchmarks, assessing real-world reasonableness, stress-testing edge cases, and identifying AI-specific hallucinations. By integrating deep code reviews for critical formulas and automated visual evidence capture via Playwright, it ensures that every calculation and UI component meets the high standards required for real estate investment and financial analysis.

主な機能

01Standardized validation reporting that generates consolidated Markdown evidence for project audits.

02Visual evidence documentation through automated Playwright screenshots of UI and report outputs.

03Integrated PAL (Programmatic Abstraction Layer) code reviews for high-stakes financial logic and formula correctness.

04Systematic four-pillar validation framework: Known Answers, Reasonableness, Edge Cases, and AI Blind Spots.

05Automated stress testing using historical extremes, zero-values, and edge-case data scenarios.

060 GitHub stars

ユースケース

01Stress-testing portfolio management tools against historical market crashes and extreme data inputs.

02Verifying real estate underwriting models to ensure mathematical accuracy and defensible investment logic.

03Auditing AI-generated financial formulas to detect hallucinations or subtle logic errors before deployment.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add peterjohn1298/re-underwriting-tool validate

For use in Claude.ai and ChatGPT

Download Skill