The CCA exam is scenario-based: every question hands you a situation and four plausible-sounding fixes, and asks for the best one. You rarely need to recall a parameter — you need to apply judgement consistently. This course drills that judgement.
Format: 60 multiple-choice questions · 120 minutes · pass at 720/1000. Five domains, weighted 27% / 18% / 20% / 20% / 15% (Agentic Architecture is the heaviest — drill it first). This is independent prep built on the public blueprint; it reproduces no confidential exam content.
The 3 rules that answer ~80% of questions
When two options look right, the one that honours these rules wins.
Rule 1 — Deterministic over probabilistic when stakes are high. Financial, compliance, identity, and security operations must use hooks, schemas, or forced tool_choice — never a prompt instruction. "Add an instruction to always verify identity before a refund" sounds right and is wrong: a prompt is followed ~99% of the time, and 1% of a refund flow is unacceptable. The answer is a PreToolUse hook.
Rule 2 — Root cause over symptom patch. Tool mis-routing → fix the tool description, not a routing layer. Narrow multi-agent results → fix the coordinator's decomposition, not the subagents. Hallucinated/absent fields → fix the schema (nullable), not the retry loop.
Rule 3 — Simplest correct intervention. Don't over-engineer and don't stack complexity on a broken architecture. If "improve descriptions" and "add a classifier layer + few-shot + a new agent" both appear, the minimal fix that addresses the real root cause is the answer.
How to read a scenario question
- Identify the symptom vs the root cause. The question describes a symptom; three options patch the symptom, one fixes the cause.
- Spot the stakes. If the scenario is financial/compliance/security, eliminate every "add a prompt instruction" option immediately (Rule 1).
- Eliminate over-engineering. Cross out answers that add layers, services, or agents where a one-line fix exists (Rule 3).
- Prefer the structural fix. Schema/hook/description/coordinator changes beat "tell the model to try harder."
Distractors are engineered from a fixed set of 22 anti-patterns. If an option is one of them, it's wrong — regardless of how reasonable it sounds. The next lesson is that anti-pattern bank.
How to use this course
- The 22 anti-patterns — learn to recognise every wrong-answer template on sight.
- Scenarios & rapid recall — the 6 recurring scenarios plus a file-locations / critical-values cheat-sheet.
- Then drill the 240-question domain quiz — pick a domain, get instant scoring and an explanation for every answer.
Pair this with the conceptual depth in the CCA Foundations course.