Exam Syllabus Contents

• Section 1. Generative AI and Prompting Fundamentals (12.5%) (5)
• Section 2. Prompt Design and Construction (30%) (12)
• Section 3. Prompting Techniques and Applied Interaction (25%) (10)
• Section 4. Evaluation, Testing, and Prompt Improvement (20%) (8)
• Section 5. Responsible, Secure, and Professional Prompting (12.5%) (5)

Section 1. Generative AI and Prompting Fundamentals (12.5%) (5)

5 objectives covered by the section → 5 exam items

1.1 Generative AI Behavior (2)

Objective 1.1.1 Explain how generative AI systems produce responses (1)

1. Define generative AI and recognize common generative AI applications.
2. Recognize large language models and multimodal models as systems that generate outputs from user input and available context.
3. Distinguish between user input, model context, and generated output.
4. Recognize that generative AI produces probabilistic outputs and may produce different responses to the same or similar input.

Objective 1.1.2 Recognize fundamental limitations of generative AI outputs (1)

1. Recognize that generated information may be incorrect, incomplete, inconsistent, outdated, or fabricated.
2. Explain the concept of hallucination as generated information that is unsupported or false.
3. Recognize that fluent, detailed, or confident language does not establish factual correctness.
4. Recognize that a model may make assumptions when information or instructions are incomplete.
5. Distinguish model-generated assertions from established facts or verified evidence.

1.2 AI Models, Tools, and Constraints (2)

Objective 1.2.1 Select an appropriate AI capability or tool for a task (1)

1. Match text, image, audio, video, coding, and multimodal tasks to appropriate AI capabilities.
2. Recognize that different models and tools may produce different results for the same task.
3. Consider required input types, output types, supported features, reliability, speed, and task complexity when selecting a tool.
4. Recognize situations in which a conventional deterministic tool or process is more appropriate than generative AI.

Objective 1.2.2 Recognize model and tool constraints relevant to prompting (1)

1. Explain the basic concept of a context window and recognize that the amount of information a model can process is limited.
2. Recognize that models and tools may support different file types, modalities, integrations, or external tools.
3. Recognize that model knowledge may have freshness or coverage limitations.
4. Recognize that output length, input size, or other technical limits can affect task completion.
5. Adapt the task or select another tool when required capabilities are unavailable.

1.3 Prompt Engineering Foundations (1)

Objective 1.3.1 Explain the purpose and basic structure of a prompt (1)

1. Define a prompt as an instruction or input provided to an AI system.
2. Define prompt engineering as the practice of designing and improving prompts to produce more useful AI outputs.
3. Identify common prompt components: task, context, input data, constraints, output requirements, and examples.
4. Distinguish the main instruction from supporting information.
5. Recognize that effective prompts depend on clarity and relevance rather than prompt length.
6. Explain how relevant context and explicit requirements help an AI system interpret the intended task.

Section 2. Prompt Design and Construction (30%) (12)

12 objectives covered by the section → 12 exam items

2.1 Task Definition and Context (4)

Objective 2.1.1 Define a clear task and intended outcome (1)

1. State what the AI system should do using clear and specific instructions.
2. Use appropriate action verbs to express the requested task.
3. Define the intended result or purpose of the task.
4. Distinguish specific, actionable instructions from vague or overly broad requests.
5. Avoid unnecessary or conflicting task instructions.

Objective 2.1.2 Define the intended audience and use case (1)

1. Identify the intended reader, listener, user, or recipient of the output.
2. Specify the purpose for which the output will be used.
3. Adapt vocabulary, complexity, technical depth, and communication level to the intended audience.
4. Recognize when audience or use-case information materially affects the appropriate response.

Objective 2.1.3 Provide relevant context and input data (1)

1. Supply facts, source material, definitions, background, metadata, or other information needed to complete the task.
2. Distinguish task instructions from the information on which the task should operate.
3. Include context that materially helps the AI interpret the request.
4. Exclude unrelated information that can distract from or obscure the task.
5. Avoid assuming that an AI system knows private, local, organizational, or unstated information.

Objective 2.1.4 Handle ambiguity and missing information (1)

1. Identify ambiguous terms, references, goals, or requirements in a prompt.
2. Identify missing information that could materially change the result.
3. Determine when to request clarification rather than make an assumption.
4. State necessary assumptions explicitly when proceeding without clarification.
5. Avoid silently inventing missing requirements or facts.

2.2 Controlling the Output (4)

Objective 2.2.1 Apply constraints to control scope and content (1)

1. Specify appropriate limits on length, scope, number of results, subject matter, or level of coverage.
2. Define required inclusions and exclusions.
3. Establish boundaries on what the response should and should not address.
4. Recognize contradictory, unnecessary, or impractical constraints.
5. Avoid over-constraining a task in ways that prevent a useful response.

Objective 2.2.2 Control tone, style, and level of detail (1)

1. Specify an appropriate tone or voice for the intended purpose.
2. Specify an appropriate reading level or technical depth.
3. Control the desired degree of detail, concision, or explanation.
4. Distinguish stylistic requirements from factual or task requirements.

Objective 2.2.3 Specify output format and structure (1)

1. Request suitable formats such as paragraphs, lists, tables, checklists, templates, or structured fields.
2. Define required headings, sections, columns, labels, or ordering.
3. Request structured data formats when the output will be processed or reused.
4. Select a format appropriate to the intended use of the result.
5. Recognize when structure makes an output easier to interpret, compare, validate, or reuse.

Objective 2.2.4 Define success criteria for an output (1)

1. State the characteristics that a satisfactory response must demonstrate.
2. Define criteria such as accuracy, completeness, relevance, clarity, consistency, and compliance with requirements.
3. Use acceptance criteria, checklists, or simple rubrics when appropriate.
4. Distinguish measurable requirements from subjective preferences.

2.3 Structuring Effective Prompts (4)

Objective 2.3.1 Organize instructions and supporting information clearly (1)

1. Separate the main instruction from context, input data, examples, and output requirements.
2. Use headings, labels, sections, delimiters, or other markers to distinguish prompt components.
3. Group related information together.
4. Make clear which content contains instructions and which content should be treated as data.
5. Present complex prompts in a logical and readable order.

Objective 2.3.2 Use roles and perspectives appropriately (1)

1. Assign a professional, functional, or audience perspective when it improves task interpretation.
2. Use roles to establish appropriate priorities, terminology, or communication style.
3. Recognize situations in which role assignment provides little practical benefit.
4. Recognize that assigning an expert role does not make generated information authoritative or correct.

Objective 2.3.3 Use examples to communicate expected behavior (1)

1. Distinguish zero-shot, one-shot, and few-shot prompting.
2. Provide examples of expected inputs, outputs, classifications, transformations, or decision rules.
3. Use examples when demonstration is clearer than additional descriptive instructions.
4. Ensure that examples are consistent with the stated task and requirements.
5. Recognize that misleading or unrepresentative examples may influence the output incorrectly.

Objective 2.3.4 Create reusable prompt templates (1)

1. Distinguish fixed prompt instructions from information that changes between uses.
2. Use clearly named placeholders or variables for changing information.
3. Design templates that support consistent execution of recurring tasks.
4. Include sufficient instructions for users to populate template fields correctly.
5. Recognize when a template needs to be adapted for a different task or context.

Section 3. Prompting Techniques and Applied Interaction (25%) (10)

10 objectives covered by the section → 10 exam items

3.1 Conversational and Multi-Step Prompting (3)

Objective 3.1.1 Manage relevant context across a conversation (1)

1. Recognize that earlier messages can influence later model responses.
2. Refer clearly to earlier instructions, decisions, source material, or outputs.
3. Restate critical requirements when they may no longer be clear.
4. Update or replace previous context when requirements change.
5. Recognize when accumulated conversation history has become irrelevant, stale, or conflicting.

Objective 3.1.2 Modify or extend an output using follow-up prompts (1)

1. Ask the AI to expand, shorten, rewrite, reorganize, or otherwise transform an existing response.
2. Identify the specific part of an output that should be changed.
3. State what should remain unchanged during revision.
4. Give targeted feedback rather than requesting an unspecified new attempt.

Objective 3.1.3 Decompose complex tasks and use prompt chaining (1)

1. Break a complex task into smaller, well-defined subtasks.
2. Arrange subtasks in a logical sequence.
3. Use the output from one step as input or context for a later step.
4. Use explicit intermediate outputs when they make a complex task easier to control or verify.
5. Recognize when several focused prompts are more appropriate than one large prompt.

3.2 Source-Grounded Prompting (2)

Objective 3.2.1 Ground AI responses in supplied source material (1)

1. Provide documents, files, excerpts, or other reference material for an AI system to use.
2. Specify which source material should be used for the task.
3. Instruct the AI to distinguish source-supported information from general knowledge or inference.
4. Limit responses to supplied evidence when the task requires source-bounded answers.
5. Recognize when the available source material is insufficient to support the requested answer.

Objective 3.2.2 Establish evidence and traceability requirements (1)

1. Request citations, references, quotations, or source locations when traceability is required.
2. Require the AI to indicate when supporting evidence cannot be found.
3. Handle conflicting source material without silently choosing one version as fact.
4. Distinguish supported conclusions from unsupported claims.
5. Verify generated citations or references before relying on them.

3.3 Common Prompting Applications (3)

Objective 3.3.1 Construct prompts for summarization and transformation (1)

1. Request summaries appropriate to a defined purpose and level of detail.
2. Rewrite, simplify, expand, translate, proofread, or reformat supplied content.
3. Specify facts, terminology, meaning, tone, or structure that must be preserved.
4. Distinguish transformation of supplied information from creation of new factual content.

Objective 3.3.2 Construct prompts for extraction and classification (1)

1. Extract specified facts, entities, values, attributes, or fields from supplied content.
2. Define categories and decision rules for classification tasks.
3. Specify the required output structure for extracted or classified information.
4. Define how missing, unknown, uncertain, or ambiguous values should be represented.
5. Prevent unavailable information from being silently invented.

Objective 3.3.3 Construct prompts for ideation, comparison, analysis, and planning (1)

1. Generate multiple ideas or alternatives according to defined requirements.
2. Compare alternatives using specified factors or criteria.
3. Request outlines, plans, recommendations, or possible courses of action.
4. Apply practical constraints to generated ideas or plans.
5. Distinguish exploratory suggestions from conclusions that require factual evidence.

3.4 Multimodal Prompting (2)

Objective 3.4.1 Construct prompts for image generation and transformation (1)

1. Describe the main visual subject and intended purpose.
2. Specify relevant setting, composition, perspective, style, mood, or visual characteristics.
3. Define important elements that should be included, excluded, preserved, added, removed, or changed.
4. Use iterative instructions to refine or modify generated images.
5. Recognize limitations involving text rendering, spatial precision, visual consistency, and exact reproduction.

Objective 3.4.2 Use multimodal inputs effectively (1)

1. Combine written instructions with images, screenshots, documents, audio, or video.
2. Request description, extraction, comparison, classification, summarization, or analysis of multimodal content.
3. Direct the AI system to relevant elements or regions of supplied material.
4. Distinguish observable information from interpretation or inference.
5. Recognize when multimodal content is unclear, incomplete, or insufficient for a reliable conclusion.

Section 4. Evaluation, Testing, and Prompt Improvement (20%) (8)

8 objectives covered by the section → 8 exam items

4.1 Evaluating AI Outputs (3)

Objective 4.1.1 Evaluate an output against prompt requirements (1)

1. Check whether the requested task has been completed.
2. Check compliance with required scope, constraints, audience, tone, and format.
3. Identify missing required information.
4. Identify irrelevant or unnecessary content.
5. Compare the output against defined success criteria or acceptance requirements.

Objective 4.1.2 Verify factual, numerical, and source-based claims (1)

1. Identify claims whose correctness matters to the task.
2. Compare important claims with reliable supplied or external sources.
3. Verify calculations, dates, quantities, quotations, citations, and references when appropriate.
4. Identify fabricated or unsupported information.
5. Distinguish plausible generated content from verified information.

Objective 4.1.3 Identify uncertainty, unsupported assumptions, and inconsistencies (1)

1. Identify conclusions based on unstated or unsupported assumptions.
2. Recognize when available information is insufficient for a confident conclusion.
3. Identify contradictions within a generated response.
4. Identify inconsistencies across repeated outputs or related responses.
5. Request qualifications or additional information when uncertainty materially affects the result.

4.2 Diagnosing and Improving Prompts (3)

Objective 4.2.1 Diagnose common causes of weak AI outputs (1)

1. Identify vague, ambiguous, incomplete, or conflicting instructions.
2. Identify missing, excessive, or irrelevant context.
3. Identify unsuitable constraints or output requirements.
4. Identify examples that conflict with or misrepresent the intended task.
5. Distinguish prompt-design problems from limitations of the selected model or tool.

Objective 4.2.2 Refine a prompt based on an identified problem (1)

1. Clarify instructions that were misunderstood.
2. Add missing information or context.
3. Remove distracting or conflicting information.
4. Adjust constraints, examples, or output requirements to address the observed failure.
5. Preserve prompt elements that already produce the intended behavior.

Objective 4.2.3 Compare prompt variants systematically (1)

1. Test alternative prompts against the same task.
2. Compare outputs using the same evaluation criteria.
3. Change relevant prompt elements deliberately rather than making unrelated changes simultaneously.
4. Recognize normal variation in probabilistic outputs.
5. Select the prompt that most consistently satisfies the defined requirements.

4.3 Testing and Workflow Reliability (2)

Objective 4.3.1 Test prompts across representative inputs and edge cases (1)

1. Use multiple test inputs rather than judging a reusable prompt from a single example.
2. Include typical, unusual, incomplete, and boundary cases where relevant.
3. Identify failures that occur only for certain kinds of input.
4. Determine whether prompt behavior remains sufficiently consistent across expected use cases.
5. Revise and retest prompts when testing reveals systematic weaknesses.

Objective 4.3.2 Use AI appropriately within a human-AI workflow (1)

1. Identify workflow steps that are suitable for AI generation or analysis.
2. Identify steps better handled by deterministic tools or human judgment.
3. Pass necessary information accurately between workflow stages.
4. Include review or verification checkpoints before important outputs are used.
5. Recognize that AI-generated work can be an intermediate product rather than the final decision or deliverable.

Section 5. Responsible, Secure, and Professional Prompting (12.5%) (5)

5 objectives covered by the section → 5 exam items

5.1 Privacy and Data Protection (1)

Objective 5.1.1 Protect sensitive information when using generative AI (1)

1. Recognize personal, confidential, proprietary, financial, authentication, and other sensitive information.
2. Avoid submitting protected information when authorization or appropriate safeguards are absent.
3. Follow organizational requirements governing approved AI tools and permitted data.
4. Consider how prompts, uploaded files, and generated outputs may be stored or processed.
5. Apply data minimization by supplying the information necessary for the task without unnecessary sensitive details.

5.2 Prompt and AI Security (1)

Objective 5.2.1 Recognize security risks associated with AI prompts and untrusted inputs (1)

1. Define prompt injection and recognize attempts to override or manipulate intended instructions.
2. Recognize indirect prompt injection embedded in documents, webpages, messages, or other external content.
3. Distinguish trusted instructions from untrusted content being analyzed.
4. Avoid exposing passwords, API keys, credentials, secrets, or restricted information.
5. Recognize increased risk when an AI system can access external tools, data, or actions.
6. Recognize that prompt wording by itself is not a complete security control.

5.3 Fairness and Responsible Content (1)

Objective 5.3.1 Recognize and reduce harmful or biased AI-assisted outputs (1)

1. Identify stereotypical, discriminatory, misleading, or unfair content.
2. Recognize when prompts themselves introduce biased assumptions or framing.
3. Request balanced treatment or additional perspectives when appropriate.
4. Avoid using generated content in ways that create foreseeable harm.
5. Review generated material before publishing or distributing it to others.

5.4 Legal, Intellectual Property, and Transparency (1)

Objective 5.4.1 Recognize legal, intellectual-property, attribution, and disclosure considerations (1)

1. Recognize that prompts, supplied material, and generated content may involve copyright, licensing, ownership, confidentiality, or attribution requirements.
2. Follow applicable organizational, contractual, academic, and professional rules governing AI use.
3. Avoid presenting third-party content as original when attribution is required.
4. Recognize situations in which AI-generated or AI-assisted work should be disclosed.
5. Recognize that legal or regulatory requirements may differ by jurisdiction or use case.

5.5 Human Oversight and High-Risk Use (1)

Objective 5.5.1 Apply appropriate human oversight to consequential AI-assisted tasks (1)

1. Recognize situations in which errors could have significant legal, financial, medical, safety, employment, educational, or other consequences.
2. Identify decisions that require qualified human or specialist judgment.
3. Avoid delegating consequential decisions entirely to generative AI.
4. Escalate uncertain or specialist matters to an appropriately qualified person.
5. Maintain human accountability for decisions and actions taken with AI assistance.