# Practical Prompt Engineering: Becoming a Prompt Specialist > An online, self-paced AI course from ARFA, the TIBLOGICS AI Academy. Intermediate level, about 19.5 hours, in English and French, with a verifiable certificate. Web page: https://tiblogics.com/learning-box/practical-prompt-engineering ยท All tracks: https://tiblogics.com/learning-box.md ## Key facts - Level: Intermediate - Time: about 19.5 hours including hands-on work (6 modules, 25 lessons, 6 labs); self-paced, no deadline - Price: $487 one time for lifetime access to this track, or every track for $89 a month (cancel anytime) - Certificate: TIBLOGICS Certified Prompt Specialist, with a public verification page - Languages: English and French ## About Anyone can get a good answer from AI now and then. A prompt specialist gets good answers reliably, on work that matters, and can show why. This track teaches you to see a prompt as one part of a system: the inputs, the model, your review, where the output goes and how the system learns. You will learn how models actually read your prompt, build prompts and prompt chains that hold up, and use critic prompts (pre-mortems, red teams, the sceptical expert, assumptions and second-order effects) so the AI uncovers your blind spots instead of just agreeing with you. Then you will test prompts like an engineer: test sets with edge and adversarial cases, pass/fail checks, rubrics, regressions and prompt injection. Every lesson has prompts you run on the page and Studio tools to compare strong and weak prompts side by side. The labs are done on the platform, with one built in a free Claude or ChatGPT account. The capstone is a tested, documented prompt system for your own work. Assumes you already use AI tools regularly. Skills this track builds also appear in prompt engineering courses from model providers and in AI practitioner certifications. This track is independent: it is not affiliated with any vendor and is not official exam preparation. Who it is for: Regular AI users in any role who want consistent, dependable results, want AI to challenge their thinking rather than flatter it, and want to become the person their team trusts with prompts. ## What you will be able to do - Explain how models read a prompt and write prompts that close the gaps a model would otherwise fill with the average - Map a task as a system and design prompt chains and reusable templates around it - Use critic prompts to make AI uncover assumptions, risks and second-order effects you cannot see - Build test sets, pass/fail checks and rubrics, compare variants and catch regressions - Apply specialist techniques: few-shot examples, structured outputs, reasoning and self-review, documents and data - Run a versioned, documented prompt library, stay current as models change, and prompt ethically ## Curriculum ### Module 1: How Models Read Your Prompt Understand what a language model actually does with your prompt, why vague prompts produce average answers, the eight parts of a strong prompt, and how to tell a strong prompt from a weak one at a glance. - What a model does with your prompt (24 min) - Why vague prompts get average answers (23 min) - The anatomy of a strong prompt (26 min) - Strong and weak, side by side (25 min) - Module quiz ### Module 2: Prompting as a System See a prompt as one part of a system with inputs, a model, your review, a destination and a feedback loop. Map a task before writing, break big tasks into prompt chains, and build reusable templates that manage context on long work. - A prompt is one part of a system (24 min) - Map the task before you write the prompt (25 min) - Decomposition and prompt chains (27 min) - Templates, variables and context for long tasks (26 min) - Module quiz ### Module 3: Making AI Tell You What You Can't See Recognise sycophancy and why models tend to agree with you. Use critic prompts (steelman, pre-mortem, red team, the sceptical expert, assumptions and second-order effects) to surface blind spots, ask for questions before answers, and stop leading the model. - Why models agree with you (24 min) - Critic prompts: steelman, pre-mortem and red team (27 min) - The sceptical expert and hidden assumptions (26 min) - Questions before answers (24 min) - Module quiz ### Module 4: Testing Prompts Like an Engineer Stop trusting a prompt because it worked once. Build test sets with edge and adversarial cases, write pass/fail checks and rubrics, compare variants and catch regressions, ask AI to attack your prompt and then check the checker, and understand prompt injection. - Build a test set (26 min) - Pass/fail checks and rubrics (25 min) - Comparing variants and catching regressions (25 min) - Let AI attack your prompt, then check the checker (27 min) - When your prompt reads the world: untrusted input, tools and skill files (25 min) - Module quiz ### Module 5: Specialist Techniques Use few-shot examples well, get reliable structured outputs such as tables and JSON, control role and style, prompt for reasoning and self-review, work with documents, data and images, and recognise when prompting is the wrong fix. - Few-shot examples done well (25 min) - Structured outputs: tables and JSON (25 min) - Role, style and reasoning (27 min) - Documents, data, images, and when prompting is the wrong fix (26 min) - Module quiz ### Module 6: Your Prompt Practice Build and version a personal prompt library, document prompts so colleagues can rely on them, stay current as models change, use prompts ethically, and assemble a portfolio that shows you work as a prompt specialist. - A personal prompt library (24 min) - Documenting prompts colleagues can rely on (24 min) - Staying current, and prompting ethically (26 min) - Your portfolio as a prompt specialist (24 min) - Module quiz ## How you are assessed - A quiz after each module (80% to pass, retakes allowed). - A timed final exam: 35 questions in 50 minutes, 75% to pass, 90% for distinction, 3 attempts. - A capstone project reviewed by a person against a published rubric (70% to pass). - Each certificate has a public page at https://tiblogics.com/certificates/ that anyone can open to check it.