Prompt Engineering
Prompt engineering is the skill of getting reliable, repeatable behaviour out of a language model by designing its instructions, context, and output format. It is assessed by whether someone can diagnose why a prompt fails and fix it systematically — not by whether they know clever phrasings.
Also called: prompting, LLM prompting, prompt design
What is prompt engineering?
Prompt engineering covers the design of instructions, context, examples, and output constraints given to a language model, plus the iteration loop used to make model behaviour reliable across many inputs rather than impressive on one.
Almost every AI product failure traces back to a prompt that worked on five hand-picked examples and collapsed on the sixth hundred. The people who can debug that gap are rare, and they are almost never identifiable from a job title.
What do the levels of prompt engineering mean?
Four levels, each defined by observable behaviour rather than by years. This is the definition employers set their bar against, and the one every proof is scored to.
- L1Aware
Writes clear single-shot prompts and can tell a good output from a bad one, but fixes failures by rephrasing until something works.
Evidence: Working prompts for well-defined, single-step tasks.
- L2Working
Uses structure deliberately — output formats, examples, explicit constraints — and can explain why a given prompt fails on a given input.
Evidence: A prompt that holds up across a varied input set, with the reasoning for each design choice.
- L3Independent
Decomposes unreliable tasks into verifiable steps, chooses between prompting, tool use, and retrieval on the merits, and builds a small evaluation set before changing anything.
Evidence: A repaired prompt system plus the test set used to prove the repair held.
- L4Leading
Sets prompting standards others follow, designs the evaluation harness the team trusts, and can predict which failures are prompt-fixable versus model-limited.
Evidence: Reusable prompt architecture and an eval suite adopted by other people.
| Level | Label | People proven here |
|---|---|---|
| L1 | Aware | 0 |
| L2 | Working | 1 |
| L3 | Independent | 1 |
| L4 | Leading | 0 |
How is prompt engineering assessed?
We give a prompt that visibly fails on a set of inputs and ask the candidate to diagnose and repair it. Scoring rewards a correct causal diagnosis, a structural fix (constraints, output format, decomposition, few-shot selection) over a cosmetic reword, and an explicit statement of how they would know the fix worked. Candidates may use any AI tool — using one well is part of the skill.
Which roles require prompt engineering?
| Role | Level required | Status |
|---|---|---|
| AI Builder | L3 Independent | Required |
| AI Engineer | L2 Working | Nice to have |
| AI Product Manager | L2 Working | Nice to have |
How do you prove prompt engineering?
Prompt Engineering: questions people ask
- Is prompt engineering still a real skill?
- The trivial part — knowing magic phrases — stopped mattering as models improved. The durable part got more valuable: decomposing an unreliable task into verifiable steps and proving the fix worked. That is what we assess.
- Can I use AI tools during the assessment?
- Yes, explicitly. Using AI well is the skill being measured. The rubric scores your diagnosis and your verification method, neither of which improves by pasting the brief into a chatbot.
- What level do employers usually require?
- Level 2 for roles where prompting supports other work, Level 3 for anyone owning an AI feature in production. Level 4 is normally only required for the person setting standards across a team.