Start here
Input to paste: (1) the workplace procedure text, unchanged; (2) role and setting, such as job title, site type, and whether this is day one or after the first week; (3) quiz size, or leave blank for 10; (4) plain-language level; (5) optional topics to weight.
Where to paste: an ordinary AI chat that accepts text. Paste the prompt after this material. No terminal or account setup is needed.
How to check the output: every question names a procedure label P1, P2 and a short "why" quoting your own wording. If a line cannot be matched back to your text, drop the question. Confirm the answer key lists letters only, and that gap flags stay out of the scored bank. Source reviewed; runtime not tested.
What this prompt does
It builds a draft quiz from your text only. It maps procedures, writes recall, scenario and sequencing questions, labels each wrong option with the move used, keeps the answer key on its own page, and lists unanswered procedure steps as questions for the trainer.
| Section | Reader use |
|---|---|
| 1 Source map | Labels procedures P1, P2 with support phrases |
| 2 Question bank | Numbered questions with A–D options and a "why" line |
| 3 Distractor rules | Names the move behind each wrong option |
| 4 Answer key | Letters plus defensive "why" lines |
| 5 Gap flags | Missing steps as trainer questions, not quiz content |
Permissions and limits
The prompt states the model has no access to your systems, files or staff and cannot look anything up. Prerequisites: procedure text in hand and a trainer who approves the quiz. It gives no legal, safety-certification or regulatory interpretation, and it does not deliver training or contact anyone. Keep real names, customers and site details out; every example stays fictional. If your text supports fewer questions than requested, the final count is reported instead of padded.
FAQ
Can I use a quiz directly with new hires? No. It is a draft for a human trainer to approve. Approve answers against your wording first.
What if my procedure text is thin? The prompt skips question types, or drops to the count your text supports, and moves passing mentions to gap flags rather than inventing steps.
Attribution
Source: TokRepo original prompt, CC BY 4.0. Reference context: ChatGPT release notes. External reference material retains its own rights.
Complete reusable prompt
You are helping me turn a workplace procedure handout into a new-hire orientation quiz. You only have the text I paste; you have no access to our systems, files or staff, and you cannot look anything up. Do not add facts, rules, phone numbers or steps that are not in my text. You are preparing a draft quiz for a human trainer to approve, not delivering training and not contacting anyone.
Input I will paste:
- PROCEDURE TEXT — the full list of workplace procedures (or the sections I name). Paste it as-is.
- ROLE AND SETTING — job title, location type (for example, front desk, warehouse floor), and whether the quiz is for day one or after the first week.
- QUIZ SIZE — number of questions I want, or leave blank for my default of 10.
- LANGUAGE LEVEL — plain-language level I should aim for, for example simple sentences for readers new to the industry.
- OPTIONAL — topics I want weighted more heavily.
What to produce:
Section 1 — Source map. List the distinct procedures you found, each with a short label and the exact phrase or sentence from my text that supports it. Number them P1, P2, and so on. If two procedures overlap or contradict each other, say so here instead of guessing which one wins.
Section 2 — Question bank. Write the number of questions I asked for (default 10). For each question give: question number; which procedure label it tests; the question text; four answer options A–D; the correct letter; and a one-line 'why' naming the exact words from my text that make that answer correct. Mix these three types roughly evenly: (a) direct recall, (b) applied scenario ('A customer asks X before opening. What do you do first?'), (c) sequencing or priority. Applied and sequencing questions must still be answerable purely from my text.
Section 3 — Distractor rules. Every wrong option must fail for a stated reason, not just sound wrong. Use only these distractor moves: swapping the order of two steps, naming a real item from the text but attaching it to the wrong stage, dropping a required confirmation, or substituting a plausible action my text does not mention. Label each distractor with the move used. Never write a wrong option that could be correct in a situation my text allows, and never write 'all of the above' or 'none of the above'.
Section 4 — Answer key on its own page, letters only in a compact list, followed by the 'why' lines so a trainer can defend every answer at a glance.
Section 5 — Gap flags. List any procedure a new hire would need but my text never states (for example, who to notify, what counts as an emergency, where to record something). Mark these as questions for the trainer, not quiz content. Do not invent answers to fill them.
Uncertainty handling: if my text is too vague to support a question of a type I asked for, skip that type and tell me instead of smoothing over it. If a procedure appears only once in a passing mention, do not build a scored question on it; put it in the gap flags. If I give fewer procedures than questions requested, stop at the number my text can genuinely support and tell me the final count.
Boundaries: no advice beyond what my text states; no legal, safety-certification or regulatory interpretation; no real names, real customer details or identifying examples. Keep every example fictional and generic. Flag, rather than resolve, anything a qualified person should confirm.
Review checks before you finish:
- Re-read each correct answer against my original wording. If the words do not clearly support it, change the question or drop it.
- Confirm no two questions test the same procedure in the same way.
- Confirm no distractor is a step my text actually requires in another order.
- Confirm every scenario example contains no real person, client or site.
- Confirm Section 5 changes nothing in the scored bank.
Worked example (fictional, to show shape only): Input excerpt: 'Front desk staff unlock the side door at 7:00 a.m. Before unlocking, check the entry log for any overnight delivery notes. If a delivery is listed, place it in the secure cupboard before opening the main doors.' A direct-recall item might ask what is checked before unlocking; an applied item might describe a listed delivery and ask what happens before the main doors open. A good distractor swaps the cupboard step to after opening, or drops the log check. A gap flag might be: who signs the entry log, which the text never says.
Output format: Markdown headings for Sections 1–5, numbered questions, plain letters for options, then a one-page answer key. Keep sentences short and at the language level I supplied.
References and reuse
Original TokRepo prompt · CC BY 4.0. Reference documents retain their own rights.