CAIN Certified Agentic Security Practitioner
Design and review a runtime control layer that decides every agent action before it runs, fails closed, and leaves evidence.
Who it's for
Security engineers, AppSec and platform teams who must let AI agents act without losing control.
Why it's in demand
AI skills now appear in more than a quarter of G7 cybersecurity job postings, and hands-on experience with AI agents is the hardest entry-level competency to find. The OWASP Top 10 for Agentic Applications gave the field a shared vocabulary in December 2025. Sources →
Exam
- Requires your Novice certificate ID
- 20 questions · 30 minutes
- 85% (17/20) to pass · server-graded
- Certificate ID C42-CCAS-…
Modules
1The execution channel: why reading prompts is not enough14 min✓2Threat modeling with the OWASP Agentic Top 1016 min✓3Fail-closed decisions: UNKNOWN and ERROR never allow12 min✓4Capability tokens and delegation that only shrinks15 min✓5Evidence-backed security reviews13 min✓
Built on this code
cain/mcp_proxy.pykeystone/bundle.pykeystone/family_verify.pyplatform-gateway/adversarial_engine.pyplatform-gateway/cain_trust_kernel.pyplatform-gateway/security_context.py