"Prompt engineering" as a career concept peaked in 2023. What replaced it in production engineering is less glamorous but more durable: applied judgment about when prompts work, when they break, and why. The CCA-F tests the latter.
What most prompt engineering curricula get wrong
Most prompt engineering courses teach you how to write prompts that produce impressive outputs in demo conditions. The techniques are real — chain-of-thought, role specification, few-shot examples — but they're taught as recipes for maximizing single-response quality, not for building systems that work reliably at scale.
Production systems fail in specific ways that demo-focused training doesn't cover:
- System prompts that work beautifully for typical inputs and fail silently on edge cases
- Over-specified prompts that constrain Claude so tightly it can't adapt to slight input variations
- Structured output schemas that Claude mostly follows — except in 2% of cases that corrupt your database
- Prompt length that looks fine in development but degrades performance when context fills with tool results
What the CCA-F tests in this domain
The Prompt Engineering & Structured Output domain (18% of the exam, 11 questions) focuses on diagnostic and remediation scenarios. You're given a system that's exhibiting a specific failure and asked to identify the root cause and the correct fix.
Scenarios in this domain typically involve:
- Structured output reliability — a system that extracts JSON from Claude's response with a high failure rate. What's the most likely cause, and what schema or prompt change fixes it?
- Instruction conflict — a system prompt and a user message that give Claude contradictory instructions. Which takes precedence, and how should the system be redesigned?
- Few-shot contamination — few-shot examples that inadvertently teach Claude a pattern that makes most outputs better but specific edge cases much worse.
- Over-specification — a verbose system prompt that's producing worse outputs than a simpler one. Why, and how do you fix it without losing the intent?
What "passing" this domain means
A strong score in Prompt Engineering on the CCA-F means you've internalized how Claude reasons about its context — not just what outputs look impressive. It means you can hand a system to another engineer and explain why the prompt is structured the way it is, what assumptions it makes about inputs, and where it will fail.
That's the difference between prompt writing as a skill and prompt engineering as a discipline. The CCA-F tests the latter.