Early Access · All certifications are currently free. Learn more
Scientia Ex Machina exists to build trust in the AI skills ecosystem. We provide rigorous, independent certification for AI agent skills - ensuring that when an AI tool says it can do something, it actually can. Safely. Reliably. Verifiably.

Our Vision

A world where every AI skill is independently verified before deployment.

AI skills are proliferating across Claude, Cursor, OpenClaw, MCP, and dozens of platforms. Nobody independently verifies they work, they are secure, or they perform. We change that.

The SSL Analogy

We are to AI skills what SSL certificates are to websites. The trust layer the ecosystem needs. Just as you would not enter your credit card details on a website without that padlock icon, you should not deploy an AI skill without independent verification that it works, it is secure, and it performs.

Leadership

Jason La Greca

Co-Founder

Former teacher turned product leader who saw a gap in the AI ecosystem that nobody was filling: independent, rigorous trust verification. Drawing on years of experience in education and product development, Jason set out to build the certification authority the AI skills world needs.

David Kellermann

Co-Founder

Former UNSW academic and now Microsoft Industry Lead for Education. David brings deep expertise in educational technology, institutional governance, and the intersection of AI and learning. His work at the frontier of education and enterprise AI grounds SXM in the rigour that institutional adoption demands.

Three Pillars of Certification

Functional Verification 40%
Security Audit 35%
Performance Benchmarking 25%

🔧 Functional Verification (40%)

Does the skill do what it claims? We run comprehensive test scenarios across diverse inputs, edge cases, and adversarial conditions to validate that skills perform as intended.

🔒 Security Audit (35%)

Is it safe to run? We test for prompt injection resistance, data exfiltration vulnerabilities, permission scope violations, and dependency risks.

⚡ Performance Benchmarking (25%)

Does it perform under load? We measure latency, accuracy, and consistency under production conditions. No slow or flaky skills get through.