How can a candidate take AI-assisted prompt engineering assessment?
AI assistants are only as good as the instructions they're given. The same tool, handed to two people, can produce sharp, usable work or a vague mess — and the difference is the person, not the model. A Prompt Engineering question measures exactly that skill: how well you can turn an AI assistant into usable output under real conditions.
You'll be given a realistic workplace scenario and asked to prompt an AI assistant to produce a result that actually solves it. What's being assessed is your instructions — how clearly you frame the task, what context you supply, and how you handle constraints and sensitive material — not luck, and not a prompt you polished somewhere else beforehand. This is a live, hands-on task, which is why it reflects your real capability far better than a multiple-choice quiz about prompting ever could.
Why this is a fair test
A few things are built in to make sure the score reflects your skill and nothing else:
- You see exactly how you're scored, every time. After each attempt, you get a full breakdown across all 8 parameters — not a mystery number. You always know where you stand and what to sharpen.
- You get room to improve. With 5 attempts and only your final one counting, early attempts are yours to learn from. You're rewarded for iterating well, not for getting it perfect on the first try.
- The score is your own. The assistant won't coach you or hint at what it's looking for. That's deliberate — it guarantees the result measures your prompting, not advice you were handed.
Before you start
The pre-test screen shows the full test: its sections, how many questions each has, and the time allotted. It also gives instructions for each question type in the test. For a Prompt Engineering section, keep these in mind:
- You get 5 attempts — this is fixed.
- You'll see an AI evaluation score after every attempt, broken down by parameter.
- Only your final attempt is scored, so use your earlier attempts to improve.
- Aim for 5/5 on every parameter.
Select Start Test when you're ready. Once you click on Start Test, the below screen appears.
A guided primer before you begin
When you reach the Prompt Engineering section, an introduction screen appears before the section timer starts. Take a moment with it — it walks you through exactly how this question type works so nothing catches you off guard once you're in. You'll see:
- The interface — where the question, resources, and chat live, and how they work together.
- How to write and refine your prompt — using the chat interface to solve the question.
- How to review the question and resources — the questions panel on the left, which holds the brief, instructions, and available resources.
- How to attach resources — using the + Add icon to bring reference material into your prompt.
- The Prompt Quality Score — the 8-parameter scorecard you receive after each response, with the aim of reaching 5/5 on every parameter.
- Improving across attempts — you have 5, and the score after each one shows you what to fix.
- The character limit — each prompt has a limit (15,000 characters); the counter shows what's left, and attached resources don't count toward it.
- The final attempt — only your last attempt is scored, so use the earlier ones to refine.
- Definitions of all 8 parameters — so you know exactly what each one means before you start.
This screen is your chance to get familiar with everything below. When you're ready, start the section.
The workspace, at a glance
When the section starts, you land on the Formulate Your Prompt screen. It has three parts:
- Left panel — the scenario brief at the top, and a Resources list below it. Each resource has a preview (eye) icon to read it and a + icon to attach it to your prompt. You can expand this panel to fullscreen or collapse it to give the chat more room.
- Center — a short reminder of how the question works: review the scenario and resources, write a prompt that defines the outcome you want, and note that the AI uses only the instructions and resources you provide. It also flags that only your latest score counts when you submit.
- Composer (bottom) — where you write your prompt. It shows your Attempts Left (5/5), a character counter (0/15000), and the assistant model in use. Send with the arrow button.
At the bottom of the screen, Next Question moves you within the section, and Submit & Go To Next Section finishes it — remember, only your final attempt on each question is scored.
Bringing in the right materials
Part of the skill is giving the assistant what it needs — and nothing it doesn't. To attach a resource:
- Select the + icon next to the resource in the left panel, or
- Type @ in the composer to open the same resource picker inline.
Once attached, the resource card shows an "✓ Attached" state, and the file appears as a chip above the composer. Remove it anytime with the × on the chip before you send. Choose deliberately: some resources may contain internal or sensitive content, and judging what belongs in your prompt is part of what's being tested.
A note on the character limit
Each prompt has a character limit, shown by the counter next to your attempts (for example, 0/15000). The exact limit depends on how the question was set up. Attached resources don't count toward it. If you go over, you'll see an inline error and won't be able to send until you trim — but your typed text is never lost.
- Left panel — the scenario brief at the top, and a Resources list below it. Each resource has a preview (eye) icon to read it and a + icon to attach it to your prompt. You can expand this panel to fullscreen or collapse it to give the chat more room.
- Center — a short reminder of how the question works: review the scenario and resources, write a prompt that defines the outcome you want, and note that the AI uses only the instructions and resources you provide. It also flags that only your latest score counts when you submit.
- Composer (bottom) — where you write your prompt. It shows your Attempts Left (5/5), a character counter (0/15000), and the assistant model in use. Send with the arrow button.
At the bottom of the screen, Next Question moves you within the section, and Submit & Go To Next Section finishes it — remember, only your final attempt on each question is scored.
Reading the scenario and bringing in the right materials
The scenario brief and its reference resources sit in a panel on the left, separate from the chat. Expand it to fullscreen or collapse it to give the conversation more room.
Part of the skill is giving the assistant what it needs — and nothing it doesn't. To attach a resource:
- Select the + icon next to the resource in the left panel, or
- Type @ in the composer to open the same resource picker inline.
Attached resources appear as removable chips above the composer before you send. Attach only what actually helps — some resources may contain internal or sensitive content, and part of the task is judging what belongs in your prompt.
Writing a prompt and reading your score
When you send a prompt, you'll see the assistant's output followed by an AI Evaluation Scorecard — all 8 parameters, each with its own score, plus a total. Each scorecard is labelled by attempt number and stays in the conversation as you go, so you can see how you're progressing.
A few things to know:
- Only your latest attempt counts when you submit — a reminder is shown before your first attempt and near each scorecard.
- If you go over the character limit, you'll see an inline error and won't be able to send until you trim it — but your typed text is never lost.
- If the assistant is temporarily unavailable, you'll see a retry option. This won't cost you an attempt or affect your results.
The 8 parameters you're scored on
- Prompt Clarity — how clearly you state what you want the AI to do.
- Instruction Structure — how well you organise the task, context, requirements, format, and constraints.
- Context Sufficiency — how well you supply the information and resources the AI needs.
- Safety & Security — how well you protect sensitive or confidential information in your prompt and the output.
- Critical Thinking — how well you understand the task and balance requirements without conflicting instructions.
- Output Correctness — how accurately the output meets the task and stays supported by the material you supplied.
- Output Specification — how clearly you define what the AI should produce: format, length, tone, audience, sections.
- Example Usage — how well you use a concrete example to show the AI the wording, format, or level of detail you want.
The assistant won't coach you — and that's on purpose
If you ask the assistant to rate your prompt, tell you how to improve it, or explain what it's assessing, it will politely decline and steer you back to the task. This isn't the assistant being unhelpful. It's what keeps the playing field level: your score reflects your own prompting skill, so the result is one you can genuinely stand behind.
