CCAR-P practice › Claude Models, Prompting & Context Engineering · Question 24 of 110
Claude Models, Prompting & Context Engineering
An employee self-service assistant answers pay questions. Its HR database tool runs under a service account that can read every employee's record. The system prompt says the assistant must never reveal another employee's salary. Security asks for the strongest control against a leak. What is the most appropriate design?
Show answer and explanation
Answer: B — Scope the tool to the requesting employee's own record using their identity, so other records can't be fetched
(b) is right: a zero-tolerance rule needs enforcement outside the model, and removing the ability to fetch other records is stronger than any check on what the model says. (a) and (d) are prompt-level controls that make a leak rarer but not impossible. (c) detects leaks after they've happened rather than preventing them.
Keep going
More Claude Models, Prompting & Context Engineering questions
- A SaaS vendor runs a Claude assistant for 30 enterprise tenants. Each tenant has a…
- A vendor-research agent fetches supplier web pages, reads internal procurement files for…
- A marketplace uses Claude to assign a condition label (new, refurbished, used, for…
- A benefits-eligibility step applies six interacting conditions, and errors cluster on…
- A lender uses Claude, with a reasoning section before each answer, to draft credit-limit…