Last released Sep 20, 2026
Applied agentic-safety infrastructure: statically detects, evaluates, and gates untrusted-input-to-dangerous-capability paths in Python and JavaScript/TypeScript AI systems.