uri-eliabayev talks with avishai-zvinsky, who leads Cato Networks’ security-content group (attack protection, traffic classification, third-party integrations). Agentic coding spread from one enthusiast with a POC; the team seeded shared rules and skills from infra code and Confluence, and agents now write their own skills after sessions. The CVE-to-protection pipeline (triage, lab exploit, signature, test on historical traffic, silent mode, block) went from half a day to as little as 45 minutes as an agentic workflow, built entirely as natural-language rules and skills. Agentic attackers are real but human-driven and error-prone; protocols don’t change, but pace does, so 12-hour response windows are gone. Cato runs an agentic attack lab on Kali Linux and is building behavioral and in-line agentic defenses.
Show: hidden-layers
Concepts: agentic-vulnerability-response · ai-cybersecurity · agent-ready-codebase · ai-rd-rollout · llm-evals