Back to Certifications
ai-prompt-engineering-v1FreeAI Prompt Engineering - Certified

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

Task Framing & Objective DefinitionContext EngineeringConstraints & Instruction DesignOutput & Schema DesignExamples & DemonstrationPrompt Debugging & IterationData, Tool & Grounding InstructionsReliability, Robustness & Safe Prompting

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.

15%

Task Framing & Objective Definition

Turns a vague ask into a precise task with a clear definition of success.

15%

Context Engineering

Gives AI the information that matters without burying the task in noise.

10%

Constraints & Instruction Design

Defines boundaries, priorities and what the model should do when conditions change.

10%

Output & Schema Design

Turns AI responses into something usable by a person, workflow or system.

10%

Examples & Demonstration

Uses examples when they improve consistency - including useful edge cases.

15%

Prompt Debugging & Iteration

Diagnoses why an AI response failed and makes the next prompt materially better.

15%

Data, Tool & Grounding Instructions

Gets AI to work reliably with supplied data, documents, APIs and tools without inventing evidence.

10%

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.

Part A20%

Prompt Foundations

Judgment about what a prompt needs for a given task - not terminology.

5 questionsfrom a bank of 20~6 min
  • Task Framing1
  • Context Engineering1
  • Constraints & Instructions1
  • Output & Schema Design1
  • Examples & Demonstration1
Part B35%

Prompt Diagnosis & Repair

A task, a prompt and the response it produced. Find what went wrong and choose the repair.

9 questionsfrom a bank of 30~13 min
  • Diagnosing a Failed Response3
  • Grounding & Tool Use2
  • Robustness & Untrusted Input2
  • Context Repair1
  • Conflicting Instructions1
Part C45%

Practical Prompt Design

Write or repair the prompt yourself. Judged on what the prompt contains, not its length.

6 written scenariosfrom a bank of 26~16 min
  • 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.

1

Your written scenario answer

40-80 words, in your own words, in the time you have.

2

AI evaluator

Reports only which expected behaviours are evidenced, and which red flags appear.

3

Deterministic rubric

Code turns those observations into points, using the bank's own signal list.

4

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.

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.