Kubernetes (CKA) Skill

A full adaptive coaching system for the Certified Kubernetes Administrator (CKA) exam, packaged as a Claude Skill. Unlike a static lab list, the skill runs real timed coaching sessions, tracks your weak spots across conversations, and tells you when you are genuinely ready to sit the exam.

What the skill does

• Two-phase coaching. Phase 1 builds coverage across all five CKA domains (Cluster Architecture/Install/Config 25%, Workloads/Scheduling 15%, Services/Networking 20%, Storage 10%, Troubleshooting 30%). Phase 2 is full exam simulation — tighter time budgets, chained failures, real performance pressure.
• Hands-on timed tasks. Each item specifies weight in points, time budget in minutes, cluster context, target namespace, acceptance criteria as observable cluster state, and any forbidden shortcuts — exactly like the real exam.
• Spaced repetition. Tasks you failed — CrashLoopBackOff root cause, NetworkPolicy default-deny, PV/PVC binding — get re-asked in disguised form until you land them twice in a row.
• A mnemonic for every mistake. kubectl one-liner patterns, RBAC verb/resource matrix, etcd backup/restore order, cordon/drain with PDB — wrong answers always come with a memory trick you can recall under time pressure.
• Imperative + declarative solutions. Every solution shows the kubectl run/create/expose/set/scale/rollout one-liner and the declarative YAML generated with --dry-run=client -o yaml, so you train both muscle memories.
• Troubleshooting emphasis. 30% of the exam is troubleshooting — the skill weights sessions accordingly and drills CrashLoopBackOff, ImagePullBackOff, Pending, OOMKilled, failed rollouts, and node problems.
• HTML progress tracker. After every session the skill generates a standalone HTML file with a mastery gauge, per-domain progress bars, weak-topic chips, a session history strip, and an "EXAM READY" banner when thresholds are hit.
• Readiness signal. The skill declares you exam-ready when all domains reach 70% and you score 85%+ on 3 consecutive Phase 2 sessions — no guessing.

How it works

1. Intake — on first run the skill introduces itself, asks your level (beginner / intermediate / advanced), background, known weak areas, and target exam date. It also shows you the exam-speed setup (alias k=kubectl, export do='--dry-run=client -o yaml', export now='--grace-period=0 --force', bash completion, vi indent settings).
2. Per-task loop — one task at a time with a running timer. You work the task, then type SHOW SOLUTION or use HINT (3 escalating hints per task). The skill produces the imperative one-liner, declarative YAML, verification commands, and common gotchas.
3. End-of-session artifacts — a summary table, a running scores table, the HTML progress tracker, and a phase-transition check.
4. Phase 2 flip — when all domains are ≥ 70% and your 3-session average is ≥ 75%, Phase 2 starts automatically. Tighter time budgets, chained failures, higher bar (85%+ × 3 in a row).
5. Continuity across sessions — the skill reads prior conversation state, so your weak spots, fastest-solved patterns, and coaching persona carry forward without you having to restate them.

How to use it

1. Click ⬇ Download this Claude Skill above.
2. Import the .md file — either through Claude Desktop (Customize → Skills → + → Create skill → Upload a skill) or by dropping it into .claude/commands/ or ~/.claude/commands/ for Claude Code. Full walkthrough in the import tutorial.
3. Invoke the skill:

   /kubernetes-cka
   

   or pass your level up front:

   /kubernetes-cka intermediate, 1 year running prod kubernetes
   

4. Run a session. Come back tomorrow, next week, or a month later — the coaching picks up where you left off.

---

Quick-start prompt (no download)

Prefer a one-shot lab list without installing anything? Paste this into Claude:

Act as a CKA trainer running a hands-on lab. Produce 10 timed practice tasks I would complete using kubectl and YAML against a real cluster.

For each task specify: weight in points, time budget in minutes, cluster context, target namespace, a 2–4 sentence scenario, the required actions stated imperatively, the acceptance criteria as observable cluster state, and any forbidden shortcuts.

The 10 tasks must collectively cover: pod with resource limits and securityContext, Deployment rollout/rollback, Service exposure (ClusterIP/NodePort/LoadBalancer), Ingress with host+path rules and TLS, PV + PVC + Pod binding, RBAC with kubectl auth can-i verification, NetworkPolicy with default-deny, node cordon/drain with PodDisruptionBudget, ConfigMap + Secret as env and files, and troubleshooting one of CrashLoopBackOff / ImagePullBackOff / Pending / OOMKilled. At least 3 tasks must be troubleshooting-class.

Do not reveal solutions until I type SHOW SOLUTION or HINT. Keep 3 escalating hints ready per task. When I ask for the solution, produce: the imperative one-liner (using kubectl run/create/expose/set/scale/rollout), the declarative YAML (generated with --dry-run=client -o yaml), the verification commands, and common gotchas.

Also show me the exam-speed setup (alias k=kubectl, export do='--dry-run=client -o yaml', export now='--grace-period=0 --force', bash completion, vi indent settings).

Tips:

• The downloadable skill is the better path if you have more than one session planned — it's the difference between a static lab list and a coach that remembers your weak spots.
• The CKA rewards speed — practice imperative commands before declarative YAML and get k + $do into muscle memory.
• Drill troubleshooting the most: it's 30% of the exam and the hardest to fake.
• Use kubectl explain <resource> during practice; it's your best in-exam reference after kubernetes.io.

⚠ This skill has been tested and optimized for Claude. Results may vary with other AI assistants.