Get certified as an
AI Prompt EngineeringCertification
Anyone can ask AI a question. Prove you can direct one - framing, context, constraints, output structure, grounding and iteration - across three parts, against a published competency blueprint.
Model-agnostic, and useful in any role that works with AI - Data Analyst, Software Engineer, AI Engineer, Product Manager and beyond.
The assessment
35 min
timed assessment
20
foundations + diagnosis + written prompt design
3
assessment sections
75%
passing standard
You will be assessed on
8 competencies, scored separately
Not one number called “Prompt Engineering”. Your result reports a level for each of these, so you can pass overall and still see exactly where the gap is.
Task Framing & Objective Definition
Turns a vague ask into a precise task with a clear definition of success.
Context Engineering
Gives AI the information that matters without burying the task in noise.
Constraints & Instruction Design
Defines boundaries, priorities and what the model should do when conditions change.
Output & Schema Design
Turns AI responses into something usable by a person, workflow or system.
Examples & Demonstration
Uses examples when they improve consistency - including useful edge cases.
Prompt Debugging & Iteration
Diagnoses why an AI response failed and makes the next prompt materially better.
Data, Tool & Grounding Instructions
Gets AI to work reliably with supplied data, documents, APIs and tools without inventing evidence.
Reliability, Robustness & Safe Prompting
Designs prompts that remain useful when inputs are ambiguous, incomplete or untrusted.
Assessment standard
Every attempt draws a fresh paper - but always the same number of questions from each competency area, so nobody gets an easier test.
Prompt Foundations
Judgment about what a prompt needs for a given task - not terminology.
- Task Framing1
- Context Engineering1
- Constraints & Instructions1
- Output & Schema Design1
- Examples & Demonstration1
Prompt Diagnosis & Repair
A task, a prompt and the response it produced. Find what went wrong and choose the repair.
- Diagnosing a Failed Response3
- Grounding & Tool Use2
- Robustness & Untrusted Input2
- Context Repair1
- Conflicting Instructions1
Practical Prompt Design
Write or repair the prompt yourself. Judged on what the prompt contains, not its length.
- Framing a Vague Request1
- Context for an Audience1
- Structured Extraction1
- Grounded Work1
- Repairing a Weak Prompt1
- Designing Examples1
You must score at least 60% here, whatever your overall total.
Assessment version
ai-prompt-engineering-v1
Question bank
76 randomized items
Passing standard
75% weighted
Result bands
Distinction / Strong Performance / Certified
Result bands describe performance on this assessment. They are not a measure of employability or hiring outcomes.
Learning outcomes
These are what the public verification page reports - and each one is claimed only where your own attempt evidenced it.
Task Framing & Objective Definition
Demonstrated the ability to turn a vague request into a precise, answerable task with a stated definition of success.
Context Engineering
Demonstrated judgment about which context a task actually needs, and the discipline to leave out what it does not.
Constraints & Instruction Design
Demonstrated the ability to state boundaries, priorities and conditional behaviour that a model can act on.
Output & Schema Design
Demonstrated the ability to specify outputs a person, workflow or system can consume without rework.
Examples & Demonstration
Demonstrated judgment about when examples improve reliability, and how to choose ones that teach the intended pattern.
Prompt Debugging & Iteration
Demonstrated the ability to diagnose why an AI response failed and make the next prompt materially better.
Data, Tool & Grounding Instructions
Demonstrated the ability to keep AI answers tied to supplied data, documents and tools rather than to invented evidence.
Reliability, Robustness & Safe Prompting
Demonstrated the ability to design prompts that stay useful when input is ambiguous, incomplete or untrusted.
How the written scenarios are graded
A model reading your answer and returning “8.7 out of 10” is not a grade - it is a guess with a decimal point. So we split the job in two: the AI only observes, and the code scores.
The evaluator reports which properties it can evidence in the prompt you actually wrote - stated the objective, closed the output to a fixed set, required a null when a field is absent, defined what to do when the data cannot answer. It has no score field to write into. Points come from a fixed rubric in code, so length and polish earn nothing on their own.
Your written scenario answer
40-80 words, in your own words, in the time you have.
AI evaluator
Reports only which expected behaviours are evidenced, and which red flags appear.
Deterministic rubric
Code turns those observations into points, using the bank's own signal list.
Weighted result + competency profile
Part A 20%, Part B 35%, Part C 45% decides the 75% pass mark; the same answers roll up into a level for each competency.
Where this fits
Each credential proves something different. You do not need all of them to start applying - but together they describe an engineer rather than a score.
Applied AI & ML Associate
Understands data and classical machine learning.
AI Prompt Engineering Certification
Can direct AI to produce reliable work. You are here.
AI Work Competency
Uses AI effectively and responsibly while working.
Tech Workplace Competency
Operates effectively inside a technology team.
Weekly Ranking
Frequently asked questions
Free, 35 minutes, and a verifiable credential at the end.
This credential verifies the ability to design, diagnose and improve prompts for AI-assisted work. It is deliberately distinct from the AI-Augmented Work Simulation, which verifies AI use while completing a real job - task delegation, output verification and end-to-end workflow execution. Passing this does not evidence those.