Blog

• The Black-Box Side of AI SecurityAnthropic built white-box AI security testing. I built the black-box counterpart — a Claude Code skill that found 32 critical vulnerabilities across 17 SaaS platforms.2026-03-15