CCAO-F Cheat Sheet: Claude Certified Associate Foundations Quick Reference [2026]

This is your quick-reference cheat sheet for the Claude Certified Associate - Foundations (CCAO-F) exam. It compresses all 7 domains into the decision rules, checklists, and contrast tables you need to have cold on exam day. CCAO-F tests judgment about using Claude well at work: picking the right model and feature, catching hallucinations before they reach a client, prompting business tasks effectively, and knowing which data never belongs in a chat. Skim this page once per day in your final week, drill whichever section surprises you, and re-read the trap-answer patterns the night before you sit the exam.

Exam Quick Facts

Four logistics details worth committing to memory. First, roughly a quarter of the 60 items are multiple-response ("Select TWO" or "Select THREE") with no partial credit, so every option you pick must be defensible on its own. Second, registration runs through the Anthropic Partner Academy, which requires Claude Partner Network membership (free to join), and passing earns you a digital badge through Credly by Pearson. Third, the credential is valid for 1 year with a free non-proctored renewal assessment available before expiration; letting it lapse means a full proctored retake. Fourth, the exam covers the claude.ai product (models, Projects, Artifacts, research mode, connectors) and deliberately excludes programming and API topics, so every question can be answered from the perspective of a professional using Claude through the app. If you write code and are weighing the developer track instead, see CCAO-F vs CCDV-F.

Domain Weights at a Glance

With 60 scored questions, each weight translates into an approximate question count. Budget your remaining review hours accordingly: the top two domains alone account for more than a third of the exam.

Domain 1: Output Evaluation and Validation (21%)

This is the heaviest domain, and its core skill is treating every Claude output as a draft that must earn your trust before it leaves your screen. A hallucination is a confident, plausible statement that is factually wrong or unsupported by any source, and the exam expects you to recognize one from symptoms alone.

Hallucination Red Flags

• Highly specific numbers, dates, or statistics with no source. Precision signals confidence and reveals nothing about accuracy.
• Named studies, publications, or quotes you cannot locate. Fabricated citations read exactly like real ones until you check them.
• Details that go beyond the material you provided. If you attached last quarter's report and the summary mentions figures the report never contains, Claude filled the gap itself.
• Uniform confidence across every claim. Claude states its weakest guesses in the same fluent tone as established facts, so tone can never substitute for verification.
• Answers about events after the model's knowledge cutoff delivered without research mode or an attached source to ground them.

Validation Checklist

Run this sequence on any output that carries consequences. Verify every name, number, date, and quotation against the original source. Ask Claude directly which part of the attached document supports a specific claim, because a grounded answer can point to its evidence and a hallucinated one cannot. Cross-check externally sourced facts in an independent reference before repeating them. Watch for bias as well as error: one-sided analysis, stereotyped examples, and missing counterarguments all count as evaluation failures on this exam.

When is human review mandatory? Anything external-facing (client emails, published copy, press materials), anything with legal, financial, or health consequences, anything feeding a decision about people (hiring, performance, vendor selection), and anything in a regulated workflow. The exam rewards the answer that keeps a person accountable at those points regardless of how good the output looks.

Output formats: an inline chat response fits quick answers and short summaries; an Artifact (a separate panel holding a standalone deliverable such as a document or table) fits anything you will revise and share; a structured format such as a table or a bulleted list fits output that feeds another step in your workflow, because structure makes checking easier.

Domain 2: Workflow Integration and Solution Design (16%)

Design questions start from the task, so state your requirements before you open a chat: the deliverable, the audience, the source material, and what a successful result looks like. An answer that jumps straight to prompting without defining the task is wrong at this level.

• Keep the person at the judgment step. The reliable division of labor has Claude producing drafts and analysis while a human makes the decision and signs off on facts.
• Communicate value and limits together. When a scenario asks how to present Claude to stakeholders, the strong answer pairs a concrete outcome (hours saved per week on meeting summaries) with an honest limitation (outputs are probabilistic and require review).
• Pilot before rollout. Integrate Claude into one workflow, measure the before and after, then expand. Big-bang adoption answers lose to measured pilots.

Domain 3: Governance, Risk, and Responsible Use (15%)

Governance questions reduce to one habit: check the policy before you paste. When you are unsure whether data is sensitive, treat it as sensitive.

Two decision rules cover most scenario questions. If an output affects people's rights, money, health, or legal standing, a qualified human must own the final call. If data would cause harm were it leaked (customer information, unreleased financials, personal records), the policy check comes before any productivity benefit.

Domain 4: Prompting and Task Execution (14%)

The Business Prompt Pattern

A strong business prompt names five things: the role or context ("You are helping a project manager prepare a stakeholder update"), the task, the source material, the constraints (audience, length, tone), and the output format. "Summarize these meeting notes for the executive team in five bullet points, each naming an owner and a deadline" beats "summarize this" on every axis the exam measures.

Task Decomposition

Task decomposition means splitting a large deliverable into a sequence of smaller prompts, each producing something you can check before moving on. For a quarterly marketing report, that sequence might be: first agree on an outline, then draft one section at a time from the relevant source material, then run a final pass for tone and consistency. Decomposition is the tested answer whenever a scenario involves a multi-part deliverable, a long document, or an output where one giant prompt returned a shallow result.

Iteration

Treat the first response as a starting point and steer it with specific feedback: "cut this to half the length, make the tone more direct, and keep the three statistics" gives Claude something to act on, while regenerating and hoping does not. The exam consistently rewards targeted refinement over rerolls.

Domain 5: Product and Model Selection (12%)

Model Selection: Haiku vs Sonnet vs Opus

The exam tests trade-off judgment across speed, cost, and quality, always anchored to a business task.

Default to Sonnet and move on evidence: step up to Opus when the task genuinely demands deeper reasoning, and step down to Haiku when the task is simple and speed matters more than polish. Recommending the most powerful model for every task is a recurring wrong answer because it ignores the cost and speed sides of the triangle.

Feature Selection Triggers

Context Limitations

The context window is the amount of conversation and attached material Claude can consider at once, and long sessions eventually strain it. Three management moves cover the exam scenarios: restart a fresh chat when a long conversation starts drifting or ignoring instructions; summarize by asking Claude to condense the decisions so far and carrying that summary into the new chat; persist durable reference material into Project knowledge so no single conversation has to hold it.

Domain 6: Configuration and Knowledge Management (12%)

A Project has two configurable layers. Project instructions are standing directions applied to every chat in the workspace: the role Claude should play, your tone standards, and required output formats. Project knowledge is the set of uploaded documents (style guides, product sheets, past reports) that Claude can draw on in every conversation. Individual chats stay separate, so the exam's classic question is knowing which layer a given piece of setup belongs in.

Connectors link Claude to live sources such as Google Drive and Gmail with your permission, so reach for one when the source material lives in those tools and changes often enough that a static upload would go stale. Configuration also needs maintenance: revisit instructions when outputs start drifting from your standards, and update or remove knowledge files when the underlying documents change, because stale knowledge produces confidently outdated answers.

Domain 7: Troubleshooting and Optimization (10%)

The exam gives you a symptom and asks for the most effective fix. Learn this table.

One optimization rule ties the domain together: change one variable at a time. Adjust the prompt, or the model, or the attached knowledge, then compare results before changing anything else, because changing several things at once tells you nothing about what worked.

Trap Answers to Watch For

CCAO-F distractors are plausible behaviors that skip a safeguard or ignore a trade-off. These patterns mark an option as wrong:

• The trust-the-output trap. Any answer that treats a fluent response as verified fact. Correct answers include validation for anything factual, external, or consequential.
• The strongest-model-everywhere trap. Opus for every task ignores speed and cost. Model choice must match task complexity and stakes.
• The one-giant-prompt trap. Cramming a multi-part deliverable into a single prompt when decomposition is the tested skill.
• The paste-first trap. Sharing sensitive data because it is convenient. Governance answers put the policy check ahead of the productivity win.
• The regenerate-and-hope trap. Rerolling an unsatisfying answer instead of giving specific, targeted feedback.
• The rebuild-every-session trap. Recreating the same context by hand in every chat when a Project with instructions and knowledge would persist it.

For multiple-response items, apply these patterns to each option independently: every selection must survive on its own, and one paste-first or trust-the-output option among your picks costs the whole question. More exam-day tactics are in How to Pass CCAO-F on Your First Attempt.

Final-Week Checklist

Work through this list in your last week. Anything you cannot check honestly is your next study block.

Next Steps

You now have the compressed version of every heavily tested CCAO-F concept. For the reasoning behind any rule on this page, return to the Complete CCAO-F Guide or the Domains Breakdown, and follow the 3-Week Study Plan if your exam date is still ahead of you.

The fastest way to find your remaining gaps is timed practice. Preporato's CCAO-F practice tests include 6 full-length exams of 60 questions each, mirroring the 7-domain blueprint with explanations for every answer, plus a 500-card flashcard deck for daily drills.