Program facts
- Study fee
- Free
- subjects
- 3
- lessons
- 6
- Estimated time
- 1.5 hours
- Access period
- 30 days
- Study language
- Arabic and English
About the program
Begin by describing a task that another reviewer can understand and check. Turn that description into a prompt that preserves input qualifications and handles missing information explicitly. Practice reviewing output structure and meaning, and compare versions without confusing improvement on a sample with a guarantee of future performance. For source-supported answers, trace which passages reach the model and distinguish an answer grounded in its context from one that is correct for current use. For operation, identify the minimum data and permissions and write a handover describing tested and untested scope. Activities involve reading cases, solving exercises and comparing outputs; they require neither a paid service nor an API integration. This prepares defined specialist tasks, not model development, penetration testing or legal consulting. Complete activities without workplace data, then transfer the method to an authorized work setting after checking its actual requirements.
Practical relevance
Apply the method to internal summaries, request extraction or review of document-supported answers. Diagnose whether a defect belongs to task definition, inputs, retrieval or output generation, then provide a record that helps the process owner select a repair. Work uses fictional cases; the program does not promise a measured business result or a particular job.
Why this program
A model response can read fluently while omitting a condition or relying on an outdated source. Work therefore begins by defining the required result and how it will be checked, rather than improving style alone. The program connects prompt writing with evaluation, grounding, data protection and tool permissions.
What you will learn
- Write a task card that defines inputs, outputs, acceptance criteria and decision authority, distinguishing missing specifications from model errors.
- Build a prompt that separates instructions, context and output requirements, with explicit behavior for missing or conflicting information.
- Design a controlled comparison, calculate its results and select an improvement using task errors and sample limitations.
- Distinguish retrieval failure, failure to use context, and source-correctness failure, then specify a testable repair for each.
- Map trust and permission boundaries for a tool-assisted task and test the separation between external content and execution authority.
- Prepare purpose-limited inputs and a handover that identifies the workflow version, tests, data boundaries, and operational responsibilities.
Learning outcomes (6)
- Write a task card that defines inputs, outputs, acceptance criteria and decision authority, distinguishing missing specifications from model errors.
- Build a prompt that separates instructions, context and output requirements, with explicit behavior for missing or conflicting information.
- Design a controlled comparison, calculate its results and select an improvement using task errors and sample limitations.
- Distinguish retrieval failure, failure to use context, and source-correctness failure, then specify a testable repair for each.
- Map trust and permission boundaries for a tool-assisted task and test the separation between external content and execution authority.
- Prepare purpose-limited inputs and a handover that identifies the workflow version, tests, data boundaries, and operational responsibilities.
Who it is for
- Practitioners seeking focused practical capability in Generative AI and Prompt Engineering
- Professionals in related roles who need applicable tools and procedures
Prerequisites
- Read work instructions and simple tables, and distinguish data from inference.
Curriculum
- Subject 1 Open for free
Task specification and prompt construction
Start with a defined function, then specify inputs, outputs, checks and authority. Repair ambiguity in the task before refining the prompt. Separate task and data, define output fields and examples, and bound inference. Check structure and meaning separately.
- Subject 2
Evaluation, retrieval and grounding
Set criteria before testing, control the comparison and separate dimensions and decisive errors. A rate describes a sample; an error log directs repairs. Diagnose before editing: finding evidence, using it faithfully, checking its validity, and covering the question are distinct judgments.
- Subject 3
Tool permissions, data and handover
Prompts guide behavior; authorization controls actions. Test both and record attempted actions separately from their effects. Minimize data by purpose, verify service conditions, and hand over scope-limited results with clear review and stop responsibilities.
Subjects unlock in sequence as you progress.
Strengths
- A checkable task specification before prompt writing
- Separate retrieval errors from response-generation errors
- A documented handover identifying authority limits and residual risks
Evidence basis
Concepts draw on the primary documentation listed in the references. Teaching examples are fictional, and a provider's recommendations for its product do not guarantee the behavior of every model. Mentioning an organization does not imply that it endorsed or helped author this program.
Methodology
A model response can read fluently while omitting a condition or relying on an outdated source. Work therefore begins by defining the required result and how it will be checked, rather than improving style alone. The program connects prompt writing with evaluation, grounding, data protection and tool permissions.
Assessment
Assessments follow the program blueprint and may use module assessments, a final assessment, or a combination, under the approved attempt policy.
Assessment attempts: 2 per assessment, 7 days apart
The pass mark is announced before the program is published.
Certificate
The credential confirms completion and successful assessment; it does not represent professional licensure unless explicitly and lawfully stated.
Every certificate has a public verification code. Verify a certificate
Disclaimer:
Program questions
Do I need a paid account to study?
You can solve the cases and output-review exercises using the supplied data without sending information to an external service. Trying a live tool is optional and subject to organizational policy and the tool's terms.
Does the program require advanced programming?
This scope does not depend on building a model or training a neural network. You should understand work tasks, read simple tables, compare results and write clear instructions.
Does this authorize me to approve a system launch?
You learn to prepare evidence and a recommendation within a specialist role. Launch approval and legal, security and privacy decisions remain with authorized organizational roles; this program does not confer independent professional authority.
Does prompt wording guarantee a correct answer?
No. You practice checking outputs, handling missing or conflicting evidence and recording errors, while distinguishing improvement on a limited sample from evidence about a defined use case.
How can I practice without using workplace data?
Use the fictional cases and supplied excerpts. When transferring the method to work, obtain the necessary authorization, use only the data needed for the task and check the relevant service settings.
How do exercises differ from summative assessments?
In exercises, write specifications, prompts, test plans and a handover, then compare them with guided feedback; this is ungraded self-practice. Six knowledge checks accompany the lessons. Three module assessments each select two applied cases, one per lesson. Automated scoring measures selection of decisions and plans appropriate to the facts; it does not alone establish proficiency in a live operational system.