owasp-ai-testing

Execute the OWASP AI Testing Guide v1 across four layers to assess AI trustworthiness.

44|2|Updated Feb 5, 2026

Search Tags:#owasp #governance #risk-management #red-team #ai-testing #audits #trustworthiness

Ready to run

npx skills add https://github.com/mastepanoski/claude-skills --skill owasp-ai-testing

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: owasp-ai-testing
Source: https://github.com/mastepanoski/claude-skills/tree/main/skills/owasp-ai-testing
Command: npx skills add https://github.com/mastepanoski/claude-skills --skill owasp-ai-testing

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

AI trustworthiness is critical for reliable AI deployments. This skill provides a structured, repeatable framework to evaluate AI systems using the OWASP AI Testing Guide v1, enabling teams to identify weaknesses before deployment and during audits.

Core Features & Use Cases

44 test cases across four layers (Application, Model, Infrastructure, Data) to assess robustness, alignment, privacy, and security.
For each test, clear objectives, observable indicators, remediation guidance, and practical guidance for audits and red-team activities.
Use cases include pre-release validation, governance compliance, risk assessment, and automated test suite generation for AI-enabled products.

Quick Start

Load the OWASP AI Testing Guide and run the 44 test cases across the four layers on your AI system.

Frequently Asked Questions about owasp-ai-testing

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

FAQPage Schema

What is OWASP AI testing and how does it evaluate AI system trustworthiness?▼

OWASP AI testing is a structured framework using the AI Testing Guide v1 to evaluate AI system trustworthiness. It applies 44 test cases across four layers—Application, Model, Infrastructure, and Data—to identify weaknesses during audits and red-team exercises.

How do I run AI red-team exercises and governance audits using the OWASP AGT framework?▼

Run AI red-team exercises by collecting inputs like system architecture, model details, and risk context, then applying the 44 OWASP AGT test cases. This generates repeatable test plans, observable indicators, and audit-ready reports for governance compliance.

Can I use OWASP AI testing for pre-release validation of AI-enabled products?▼

Yes, you can use OWASP AI testing for pre-release validation. It systematically assesses robustness, alignment, privacy, and security across the Application, Model, Infrastructure, and Data layers to identify weaknesses before deployment.

What inputs do I need to provide for an AI risk assessment using the OWASP AI Testing Guide?▼

You need to provide inputs including ai_system_description, system_architecture, testing_scope, model_details, data_details, existing_controls, and risk_context to generate repeatable test plans and remediation guidance.

Best way to generate automated test suites for AI trustworthiness and compliance checks?▼

The best way is applying the OWASP AI Testing Guide v1 framework, which provides 44 structured test cases across four layers with clear objectives, observable indicators, and remediation guidance for automated test suite generation.

What are the limitations of using the OWASP AI Testing Guide for production validation?▼

The OWASP AI Testing Guide requires comprehensive inputs like system architecture and data details to be effective. Without complete existing controls and risk context information, the generated test plans and audit-ready reports may lack full coverage of trustworthiness indicators.