🧪 Labs

Homelab Architect for Cert Practice prompts

Five prompts to make hands-on practice actually stick — design a homelab on your budget, generate graded lab exercises, debug a broken lab as a learning moment, and turn the work into a portfolio project.

Tested 2026-05 Claude 4.7 OpusGPT-5Gemini 2.5 Pro #homelab#study#cert#hands-on
Honest note — A model can design a lab and explain an error, but it cannot see your machine. Always cross-check commands and quotas against current official docs — and never paste real credentials, keys, or internal IPs into a chat.

Prompts in this set

  1. 1. Design a homelab for a specific certification
  2. 2. Generate a progressive set of lab exercises
  3. 3. Debug a broken lab as a learning moment
  4. 4. Turn homelab work into a portfolio project
  5. 5. Choose between lab tooling options

1. Design a homelab for a specific certification

When you know you need hands-on practice but don't know what to actually build or run it on.

Claude 4.7 Opus (2026-05)GPT-5 (2026-05)
Act as a homelab mentor. Design a practice lab for a certification.

Certification: <CERT> (e.g. CCNA, RHCSA, CKA, AWS SAA)
Hardware I have: <RAM, CPU, DISK — or "just a laptop">
Monthly budget: <e.g. 0€ / 10€ / 30€>
Goal: pass the exam AND have something real to show.

Deliver:
1. The leanest setup that covers the exam's hands-on objectives — name the specific tools (e.g. VirtualBox, Proxmox, GNS3, kind, cloud free tier) and why.
2. What runs locally vs what should run on a cloud free tier, with the cost risk of each.
3. A 4-stage build order, from "first thing to install" to "exam-realistic environment".
4. The one mistake most people make with a <CERT> homelab and how to avoid it.

Keep it within my budget and hardware. If my hardware genuinely can't do it, say so and give the cheapest cloud alternative.
TipBe honest about hardware — 8 GB of RAM and 16 GB lead to very different designs. Re-run it if the exam blueprint changed recently.

2. Generate a progressive set of lab exercises

When your lab is built but you just poke at it randomly instead of practising the actual exam skills.

Claude 4.7 Opus (2026-05)GPT-5 (2026-05)
Create a set of hands-on lab exercises for one exam objective.

Certification: <CERT>
Objective to drill: <SPECIFIC_OBJECTIVE> (e.g. "configure OSPF", "manage LVM storage", "set up an Ingress")
My current level on it: <NEW | SHAKY | NEED_SPEED>

Produce 5 exercises, easy to exam-hard, each with:
- A task stated the way the exam would state it (terse, outcome-based — not a tutorial).
- The success criterion: how I verify it actually worked.
- A time budget a competent candidate should hit.
- A hidden "hint" line I can choose to read only if stuck.

Do NOT include the solution commands — I learn by doing, not reading. End with one "break it and fix it" exercise that mimics a real troubleshooting question.
TipAsk for the solutions in a separate follow-up message, only after you've attempted all five — keep the first reply solution-free.

3. Debug a broken lab as a learning moment

When something in the lab won't work and you want to understand why, not just paste a fix.

Claude 4.7 Opus (2026-05)GPT-5 (2026-05)
My homelab setup is broken. Walk me through it like a patient senior engineer — teach, don't just hand me the fix.

What I was trying to do: <GOAL>
Environment: <OS, TOOL, VERSIONS>
What I expected vs what happened: <SYMPTOM>
Error output (no secrets/IPs): <PASTE_SANITISED_OUTPUT>
What I've already tried: <STEPS>

Respond in this order:
1. The most likely root cause, in one sentence, and how the symptom points to it.
2. A diagnostic command or check to confirm it BEFORE changing anything.
3. The fix — and what it actually does, so I'd recognise this class of problem next time.
4. The exam-relevant lesson: if a <CERT> question hit this scenario, what would it be testing?

If there isn't enough information, tell me exactly what to collect.
TipSanitise the paste — strip real IPs, hostnames, keys, tokens. The lesson works fine on redacted output.

4. Turn homelab work into a portfolio project

When you've built something in the lab and it's just sitting there instead of strengthening your CV.

Claude 4.7 Opus (2026-05)GPT-5 (2026-05)
Help me turn a homelab build into a credible portfolio project.

What I built: <DESCRIBE_THE_LAB_OR_SETUP>
Cert / role it supports: <CERT_AND_TARGET_ROLE>
Where it'll live: <GITHUB | BLOG | BOTH>

Produce:
1. A project framing — the problem it 'solves', stated like a real-world scenario rather than "I followed a tutorial".
2. A README outline: overview, architecture (with an ASCII diagram), how-to-run, what I learned, what I'd improve.
3. Three resume bullet points derived from it — action verb, concrete tech, outcome — honest, no inflation.
4. Two questions an interviewer could ask about this project, with a one-line answer each.

Keep it truthful to what I actually built. If it's too thin to be a portfolio piece, tell me what to add first.
TipDocumentation is the project. A working lab with no README is invisible to a recruiter; a modest lab with a clear README is not.

5. Choose between lab tooling options

When you're stuck comparing virtualization or lab tools instead of actually studying.

Claude 4.7 Opus (2026-05)GPT-5 (2026-05)
Help me pick the right tool for my homelab — decisively, not a feature dump.

Decision: <OPTION_A> vs <OPTION_B> (e.g. VirtualBox vs Proxmox, GNS3 vs CML, kind vs minikube, local VM vs cloud free tier)
What I'm using it for: <CERT_OR_OBJECTIVE>
My constraints: <HARDWARE, BUDGET, OS, TIME>

Give me:
1. A one-line verdict for MY constraints — pick one, don't fence-sit.
2. The 3 trade-offs that actually matter here (ignore the rest).
3. The situation in which the OTHER option would have been the right call.
4. The hidden cost of the option you recommend (setup time, RAM, licence, learning curve).

If both are overkill for what I need, say so and name the simpler third option.
TipGive it real constraints. "I have 8 GB RAM and need to pass CCNA in 2 months" gets a sharp answer; "which is better" gets a shrug.

How to use these prompts

Each prompt has placeholders in <ANGLE_BRACKETS> — fill them in before pasting. Copy the prompt with the button, paste into Claude, ChatGPT, Gemini, or any chat-UI'd LLM.

Why "model tested" dates matter

LLMs improve and regress with every release. A prompt that worked on Claude 3.5 may need rewriting for Claude 4. The dates show when each prompt was last verified — anything older than 6 months should be re-tested before depending on it.

Found a better prompt?

Hit contact and share — we keep prompts that beat ours.