Course Test

Question 1

A platform engineer wants every LLM response that contains Kubernetes configuration to use a specific YAML fenced code block format with no surrounding prose. Where should this output format requirement be defined?

Select one answer

In each user turn, reminding the model of the format for that request

In the system prompt, so it applies to all responses without repetition

In a post-processing script that reformats the raw model output

In a Kubernetes ConfigMap that the agent reads before responding

Question 2

An SRE asks an LLM to 'optimize the resource requests for my api-gateway deployment' without providing any cluster state. What is the most likely failure mode?

Select one answer

The LLM will refuse to answer because the task is too vague

The LLM will produce a manifest with fabricated resource values that do not reflect actual usage

The LLM will call kubectl directly to fetch the current state

The LLM will produce a manifest with no resource requests at all

Question 3

Which kubectl command validates a model-generated Kubernetes manifest against the cluster API without making any changes to the cluster?

Select one answer

kubectl validate -f manifest.yaml

kubectl apply --dry-run=client -f manifest.yaml

kubectl apply --dry-run=server -f manifest.yaml

kubectl diff -f manifest.yaml

Question 4

A team uses a custom label scheme (team, managed-by, cost-center) on all their Kubernetes resources. They want an LLM to always include these labels when generating new manifests. Which prompt technique is most effective for teaching the model this convention?

Select one answer

Role prompting - assign the model the persona of a senior platform engineer

Few-shot prompting - provide two or three example manifests with the correct labels

Step-by-step decomposition - ask the model to add labels one at a time

Checklist prompting - include the label requirements in a review checklist

Question 5

What is the 'retrieval over recall' principle in the context of infrastructure prompting?

Type your answer below

Question 6

Which of the following are valid reasons to run your prompt eval set? Select all that apply.

Select all that apply

When the system prompt is changed

When the underlying model or model version is updated

When a user reports that the prompt produced unexpected output

Only when a completely new prompt is written for the first time