CCAR-F exam scenarios: all six are published, and the smallest domain frames four

The CCAR-F exam scenarios are four drawn from a bank of six — a line in the exam-facts block of the official guide, and the only statement of structure anywhere in the four-exam programme. Sixty scored items across four scenarios averages about fifteen items to a premise, which makes one scenario roughly a quarter of the paper.

The only published structure

The other three guides are not being cagey; they simply present their papers as a count of items against a set of weights, and say nothing about how those items are assembled. So on three of the four exams you learn the shape of the paper by sitting it. On CCAR-F you can know beforehand, and it is worth more than a curiosity.

One premise, a quarter of the paper

This is the part that changes how you read. If sixty items arrive as four scenarios, the average scenario carries fifteen of them, and every one of those fifteen rests on the same described system. A misread premise is therefore not a lost question; it is a lost block, and the loss is silent, because each individual item still looks answerable.

Which makes the opening minute of each scenario the highest-leverage minute on the paper. Read the system description once slowly and fix the constraints in mind — what it does, what it is not allowed to do, what has already gone wrong — before touching the first question. The general version of that skill is in how to read a Claude exam question, and it pays four times as much here as on a paper of independent items.

What the six actually are

Section 5 of the guide names all six and prints, under each, the domains it primarily frames. It gives the production context of each system — not a single question — so this tells you what you will be reasoning about, and nothing about what you will be asked:

Count down the right-hand column and the blueprint inverts. Context Management and Reliability is the smallest domain on the paper — 15%, about nine items — and it is a primary domain in four of the six scenarios, more than any other. Prompt Engineering and Structured Output is joint-second-largest at 20% and frames only two.

Which is a real study consequence rather than a curiosity. Context management carries the fewest marks of any CCAR-F domain and is the connective tissue of two thirds of the paper, so it is the one domain where weight understates how often you will need it — context management and reliability on CCAR-F covers what it actually asks.

Two sittings must share two

That is arithmetic, not policy, and the distinction matters. The guide states the structure of an exam; it does not say how a second attempt is drawn, so the honest claim is a floor on overlap conditional on the same bank, and nothing about which two scenarios repeat or whether the items inside them do. What the retake rules actually say is in failed the Claude certification, what next.

What it changes about pacing

It gives you a unit. CCAR-F publishes no per-domain time budget, so the natural split is by scenario: four scenarios in 120 minutes is thirty minutes each, and a scenario running long is a signal you can act on while three remain rather than a realisation at the end.

And unlike the developer and professional guides, whose published budgets spend the entire sitting, nothing here has already claimed your minutes — so build the reserve in from the start, at roughly twenty-seven minutes a scenario. The comparison across all four papers is in pacing on the Claude exams.

What it does not change

The blueprint still governs. The paper is weighted by domain — Agentic Architecture and Orchestration at 27%, Tool Design and MCP Integration at 18%, and so on — and those weights hold across the whole exam regardless of how the items are packaged. A scenario is a container, not a syllabus unit.

It also does not mean a scenario sits inside one domain. The opposite is closer to true: a described system invites questions about its architecture, its tool contracts, its context handling and its failure modes, which is four domains inside one premise. So study by domain, as the domain weights and our Claude certification study guide set out, and let the structure change how you read on the day rather than what you read before it.

Key takeaways

• Four scenarios from a bank of six. Stated in the CCAR-F guide, and the only published structure of the four exams.
• About fifteen items to a premise. A misread scenario costs a quarter of the paper, quietly.
• Two sittings must share at least two. Four plus four against six, if a retake draws from the same bank.
• All six are published by name. Section 5 lists them and the domains each one frames.
• The smallest domain frames the most scenarios. Context Management at 15% is primary in four of six; Prompt Engineering at 20% in two.
• Study by domain anyway. Scenarios span domains; the weights still decide where the marks are.