Owasp Ai Testing

This implements the OWASP AI Testing Guide v1 as a structured testing methodology for AI systems. You get 32 test cases across four layers: application (prompt injection, data leaks), model (robustness, alignment), infrastructure (supply chain), and data (bias, privacy). Each test includes objectives, payloads, observable indicators, and remediation steps. The skill requires system descriptions and testing authorization as inputs, and it defaults to documentation-only mode unless you explicitly approve active testing. Worth noting that this is a how-to-test guide, not a vulnerability checklist, so pair it with OWASP LLM Top 10 if you need threat identification alongside the testing procedures.