CCAR-F Context and Reliability Practice Question
A development team is building an internal document Q&A tool. They want Claude to answer strictly from uploaded policy PDFs and to decline questions the documents do not cover. Which prompt structuring technique best supports this requirement?
⚠ Common exam trap
The trap here is believing that sampling parameters or extra examples can enforce source restrictions, when only explicit delimiting plus a decline instruction defines the answering boundary.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Provide the documents inside XML tags and instruct Claude to answer only using content within those tags, declining otherwise.
Strict document-scoped answering needs two things: unambiguous boundaries around the source material and an explicit instruction about what to do when the sources are silent. Delimiting the PDFs in XML tags gives Claude a referable container, while the decline instruction supplies the fallback behavior. Sampling tweaks, example counts, and knowledge-first ordering all leave the sourcing boundary undefined.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Ask Claude to answer from its general knowledge first, then compare with the documents afterward.
Why it's wrong here
Answering from general knowledge first invites the model to produce content outside the uploaded policies, violating the strict sourcing requirement. Even if a comparison step follows, the primary answer already risks unsupported claims. The team explicitly wants responses limited to the PDFs, so leading with parametric knowledge inverts the intended constraint and undermines the tool's reliability.
- ✓
Provide the documents inside XML tags and instruct Claude to answer only using content within those tags, declining otherwise.
Why this is correct
Wrapping source material in XML tags creates clear boundaries that Claude can reference in instructions, and pairing that structure with an explicit decline rule gives a testable answering contract. This directly implements the team's requirement that answers come only from uploaded policies. It uses structural delimiting plus a behavioral constraint, which is exactly what strict document-scoped question answering demands.
- ✗
Increase the number of few-shot examples to fifty so the model learns the answering pattern.
Why it's wrong here
More examples can shape format and tone, but they do not define which content is authorized as a source. Without delimiting the documents and stating a decline rule, the model still has no enforced boundary between policy text and prior knowledge. Example volume addresses style imitation, not source restriction, so it leaves the core requirement unmet.
- ✗
Set the top_p parameter to 0.1 to make the model choose only high-probability tokens.
Why it's wrong here
Narrowing nucleus sampling makes word choice more conservative but does not tell the model where its facts may come from. A high-probability token can still be an unsupported fact drawn from training data. Sampling parameters influence fluency and diversity, not the scope of permitted sources, so this setting cannot enforce document-only answering.
About these practice questions
This CCAR-F question is part of Courseiva's 271-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCAR-F practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCAR-F exam.