[{"data":1,"prerenderedAt":31},["ShallowReactive",2],{"blog-tag-en-ai-agent":3},[4,15,23],{"path":5,"title":6,"description":7,"date":8,"tags":9},"\u002Fblog\u002Fen\u002Fk3s-mcp-ai-agent-rbac-permission-bypass","We Asked an AI Agent to 'Check a Pod,' and It Took Over the Entire K3s Cluster: What an MCP Server's Permission Bypass Teaches Us About RBAC Design","A real-world MCP permission bypass shows why Kubernetes\u002FK3s clusters need disciplined RBAC design and audit-log separation before letting AI agents operate them.","2026-09-03",[10,11,12,13,14],"k3s","kubernetes","mcp","rbac","ai-agent",{"path":16,"title":17,"description":18,"date":19,"tags":20},"\u002Fblog\u002Fen\u002Fai-agent-authentication-kubernetes-keycloak-spiffe","Don't Hand AI Agents the Keys. A Keyless Design for Authenticating MCP Servers on Kubernetes","Handing AI agents static API keys is an operating model that eventually breaks down. This article explains why static secrets hit a wall when running MCP servers on Kubernetes, and how Keycloak combined with SPIFFE\u002FSPIRE enables a 'keyless' authentication design.","2026-08-18",[10,11,14,12,21,22],"keycloak","zero-trust",{"path":24,"title":25,"description":26,"date":27,"tags":28},"\u002Fblog\u002Fen\u002Fai-agent-sandbox-kata-containers-kubernetes","AI Agent Code Isn't a \"Trusted Product\" Anymore. Kubernetes Sandbox Design Has an Answer","Code generated and executed by AI agents can no longer be treated as a trusted, reviewed product. This article explains the limits of container isolation and why Kata Containers' microVM isolation is becoming essential when designing AI agent sandboxes on Kubernetes.","2026-08-08",[10,11,29,14,30],"kata-containers","security",1792488945760]