CCDV-F Security Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Increase the max_tokens setting so Claude has more room to reason about whether input is malicious.

  Why it's wrong here

  Token limits govern output length, not security reasoning. Giving the model more space to generate text does not create a reliable injection detector and can actually give an attacker more room to steer the conversation. This does not address the authorization gap and is not a recognized prompt injection control.

• ✗

  Cache responses for identical user inputs to reduce the number of times Claude processes potentially malicious text.

  Why it's wrong here

  Caching changes how often the model is invoked, not whether an injected instruction is honored. A cached response could even replay a previously successful injection to other users if keys are not properly scoped. This does not reduce the risk of unauthorized actions and introduces additional data-isolation concerns, so it is not an effective control.

• ✓

  Delimit and clearly label untrusted user content within the prompt, instructing Claude to treat it as data rather than instructions.

  Why this is correct

  Wrapping user input in explicit delimiters with a clear statement that the enclosed content is data, not instructions, reduces the chance that Claude follows injected commands. It is not a complete boundary on its own, but it is a standard, low-cost mitigation that measurably lowers injection success and preserves the intended task behavior when combined with server-side authorization.

• ✓

  Enforce authorization for any action Claude proposes by validating the user's session and entitlements in the application backend before executing.

  Why this is correct

  Even if injection succeeds in making Claude propose an unauthorized action, server-side authorization checks the authenticated user's permissions before executing anything. This is the decisive control because it does not depend on model behavior. It converts a potential privilege escalation into a denied request, which is why it is essential alongside prompt-level mitigations.

• ✗

  Lower the temperature to zero so Claude becomes deterministic and cannot be manipulated.

  Why it's wrong here

  Temperature affects sampling randomness, not susceptibility to adversarial instructions. A deterministic model can still reliably follow an injected command every time, which is worse for security. Determinism aids reproducibility and testing, but it is not a mitigation for prompt injection and does not prevent unauthorized actions.