Examworthyexamworthy.comClaude Certified Architect - Professional cheat sheet
Anthropic
Free to share. Examworthy is not affiliated with or endorsed by Anthropic; CCAR-P and related marks belong to their respective owners.
At a glance
Format: Proctored multiple-choice and multiple-response; each item states how many responses to select
Domain weight map
Heaviest first - spend your time hereHow this exam thinks
CCAR-P tests whether you can pick the single best architectural decision for a stated situation, preferring a structural fix over a compensating control and diagnosing from what changed.
Spot the trap
Tempting wrong answers, and why they failCommon misconception
Logging every call with an alert is not equivalent to removing the tool. It is a detective control: the harmful write has already happened by the time the alert fires, and the capability itself remains.
Integration
Common misconception
Lowering temperature to zero makes outputs more consistent but does not guarantee identical results across calls, and a generated calculation still cannot be traced to a specific source row.
Solution Design & Architecture
Common misconception
A fall in cost per request does not carry through to the bill when retries multiply the requests per task, and it does not count human rework at all.
Evaluation, Testing & Optimization
Common misconception
Clear instructions and examples usually lower the violation rate a lot, but a lower rate is not the requirement. At thousands of items a week even a small residual rate publishes violating copy, and a prompt cannot guarantee zero.
Governance, Safety & Risk Management
Common misconception
Training on past approved outputs does not keep policy knowledge current: with frequent revisions the weights go stale within days, and a fine-tuned model cannot point to which clause version it relied on.
Stakeholder Communication & Lifecycle Management
Common misconception
A stronger tier does not make extra reasoning unnecessary. On multi-step calculations the most capable tier answering directly can still fall below a quality floor that a mid tier with extended thinking clears.
Claude Models, Prompting & Context Engineering
Common misconception
An instruction in CLAUDE.md is guidance to the model rather than an enforced permission, so it is a weak sole control for a security requirement, and maintainers could later revert it.
Developer Productivity & Operational Enablement
Common misconception
A larger model tier will not close an injection hole. Resistance does vary between tiers, but no tier reliably refuses every injected instruction, and while the write tools remain any successful injection still has a path to act.
Integration
Key terms
Exam-day rules
- Read the last sentence of the stem first. It tells you whether you are recommending, diagnosing or naming a cause, which changes what a correct option looks like.
- List the constraints before reading the options. A deadline, a latency target, a staff workflow that cannot take interruptions or an audit finding will eliminate most distractors on its own.
- When one option removes a cause and another watches or confirms it, prefer the removal unless the stem shows the capability is genuinely needed.
- On any regression stem, find what changed in the window before you consider tuning. An option that upgrades an unchanged component is usually the trap.
- Distrust any design where the model supplies its own scope, permission or identity. Those decisions belong in the system, enforced outside the model.
Revision schedule
- Day 1Read the exam guide and map the seven domains
- Week 1Draw two systems end to end
- Week 2Audit integration: tools, identity and retrieval
- Week 3Practise evaluation and the what-changed diagnosis
- Week 3 to 4Cover governance and the stakeholder domains