Blog topic
Plan safety, hard denies, risk levels, and why --approve is never a substitute for review. 7 articles.
Muhtalip Dede3 min readDay-2 Helm with kprompt: real helm install/upgrade plans, template and client dry-run previews, Bitnami recipes, and hard denies for uninstall-all — without replacing Helm or GitOps.
Muhtalip Dede3 min readWhy unsupervised auto-remediation is not the destination. Observe by default, Autopilot propose-only, reality anchors, and investigate → plan → approve → verify as the load-bearing loop — not a fleet of agents that apply because the model sounded sure.
Muhtalip Dede4 min readChat is a line; ops is a gated graph. Fan-out where edges are real, independent verify (not same-session soft-agree), PlanResult → approve → verify — without becoming a free-form agent fleet.
Muhtalip Dede3 min readAI SRE across kubeconfig contexts: single-context default, explicit read fan-out, per-context mutate approval, aliases, and why we refuse silent fleet --approve or uploading cluster credentials.
Muhtalip Dede2 min readPolicy is code, not LLM vibes. How kprompt’s safety engine hard-denies wipe-class intents, scores risk, forces approval, and why fail-closed is the load-bearing wall of AI SRE.
Muhtalip Dede6 min readEdge-case prompt scenarios for Kubernetes AI CLIs — wipe jokes, unscoped deletes, ambiguous resource names, missing tools, secrets reads, scale-to-zero, and --approve traps — with what kprompt does today.
Muhtalip Dede4 min readWhy natural-language Kubernetes tools need plan-before-apply, risk scoring, and hard denies — with real kprompt examples for scale, rollback, and blocked wipe prompts.