P2 · 8 questions

Claude Models, Prompting & Context Engineering — Attempt 1

CCAR-P P2 practice. Some items are “select all that apply” — they tell you how many to pick and are graded all-or-nothing. Commit before revealing, then read every explanation.

0 of 8 answered

P2 · Q1model-family selection
A team needs to route incoming support tickets into one of five categories at very high volume, where speed and per-request cost matter far more than nuanced reasoning. Which Claude model tier is the most appropriate default for this subtask?
P2 · Q2no fine-tuning of Claude
A customer says: 'We want Claude to always answer in our brand voice, so we plan to fine-tune the Claude model on our style guide.' What is the most accurate response?
P2 · Q3prompt structure (XML tags)
When a single prompt contains both a long reference document and a set of instructions, which structuring choice best helps Claude reliably separate the two?
P2 · Q4right-sizing per subtask
A pipeline first extracts fields from scanned invoices, then writes a nuanced financial summary of them. Cost is a concern. What is the most appropriate model assignment?
P2 · Q5prompt caching
You want prompt caching to cut cost on a chatbot that reuses a large fixed system prompt, but each request currently begins that system prompt with the current timestamp and cache reads are always zero. What is the best fix?
P2 · Q6context rot / lost-in-the-middle
An agent stuffs roughly 300K tokens of loosely relevant history into every call, and accuracy on facts stated early in that context steadily degrades. What is the most likely root cause?
P2 · Q7extended/visible thinkingSelect 2
Select the two situations where enabling extended/visible thinking (a higher reasoning budget) is most justified despite its token cost.
P2 · Q8structured output & schema reliabilitySelect 2
A team forces tool/JSON schema output and assumes the response is therefore trustworthy. Which two statements are correct cautions?

More P2 attempts