The exam is a practical test, like a driving test in a real car: nobody asks you to describe a three-point turn, you do one, and the examiner only looks at where the car ends up. Each task says which car to drive, and parking perfectly in the wrong car scores nothing. The drills here are practice drives: a new route every round, the same moves each time, until they feel automatic.
Exam-shaped practice
The CKS gives you 2 hours for 15-20 hands-on tasks (the Linux Foundation's "Important Instructions: CKS", checked 21 Oct 2026; Kubernetes v1.35, 67% to pass; the killer.sh simulator that comes with it has 17 per session). Each task names the host to work on: ssh <host> from the base machine (kubectl with the k alias, yq, curl and man pages are set up there, not on base), do the task, exit back to base (no nested ssh). Every task has a weight and is graded on the state you leave behind, never on the commands you typed.
How the drills run. Each round builds a fresh scenario with new names: the namespace, the profile, the image, the node all change, so you cannot answer from memory, only from skill. (simulator) The lab is one box with one cluster, so a task names a kubeconfig context instead of a host (the older exam format), and the shell starts in the wrong one on purpose. Answer files go under /opt/course/<task>/.
Three habits that score.
Go where the task says first, every task: its host on the exam, its context here. A perfect fix on the wrong cluster scores zero.
Node work is ssh work. AppArmor, seccomp, kube-bench, the manifests and Falco's log live on the node: ssh <node>, do it, exit. The static pod restarts by itself when its manifest changes.
Verify, then move on.kubectl get, auth can-i, a denied kubectl run, a re-run of kube-bench: one check per task before the next one.
Time. 6-8 minutes a task at 15-20. A task that fights you after 10 minutes is flagged and left; the block drill at the end trains exactly that triage.
The same tasks are the CKS exam mode (Progress › Certifications, or the phone's Cards › Exam): there you read a task, plan it, and compare with the reference solution.
Why it helps
Knowing each control is not the same as applying it quickly, under a clock, on the right cluster. The drills rebuild every task with new names each round, so you practise the skill, not an answer you remember, and the five-task block trains the triage the 2-hour exam needs. The same habits carry over to real security work: change the right cluster, prove the control works, then move on.
On the state you leave behind: the objects, files and flags the task asked for, checked after the exam. The commands you typed are not graded, so verify each task before you move on: a get, a can-i, a denied request or a re-run of kube-bench.
Why does every task start by naming a host (or a context)?
The exam has several clusters, and each task names the host to ssh to from the base machine, where kubectl is set up for that cluster; older exams and these drills name a kubeconfig context instead. A correct fix applied to the wrong cluster scores zero, so getting there is the first step of every task, and you exit back to base afterwards.
Which tasks need ssh to a node?
Anything that lives on the node rather than in the API: AppArmor and seccomp profiles, kube-bench, the static pod manifests of the control plane, and Falco's log. You ssh in, make the change, and exit; a static pod restarts by itself when its manifest changes.
How much time does a task get?
The Linux Foundation's exam instructions say 15-20 tasks in 2 hours, so 6-8 minutes each (its killer.sh simulator has 17 in a session). A task still fighting you after 10 minutes is flagged and left for the end, because the points are the same for an easy task and a hard one of the same weight.
What is the difference between the drills and the exam mode?
A drill builds the task in the simulated cluster and checks your work. The exam mode shows the same tasks to read and plan, then the reference solution to compare with, which works on the phone too.
In an interview Mid
How do you prepare for a hands-on exam like the CKS?
By doing the tasks, not reading about them: practise each kind of task until the steps are automatic, under a time limit, on fresh names every time. On the exam, go to the task's host first, use ssh for node work, verify every task before moving on, and flag anything that takes longer than about 10 minutes.
Also asked: Which CKS domain is the hardest for you, and why? · How do you verify a security control actually works? · How do you decide what to skip under time pressure?