Teach prompting like a subject, not a trick.
iCRAFT was built in the classroom. Six named engines give students a shared vocabulary, the formulas give them scaffolding, and the rubric gives you something you can actually grade. Here's everything you need to run it.
How to teach iCRAFT in class
1. One engine per session
Don't teach all six at once. Each engine follows the same 4-beat lesson: the question it asks → the formula → a bad-vs-better example → hands-on practice in the Lab. 45–60 minutes per engine.
2. Practice before theory
Open each session by having students write a prompt cold, score it in the Practice Lab, and keep the score. Teach the engine, then have them retry. The score jump is the lesson.
3. Shared vocabulary
Once students know the engine names, feedback gets fast: "This needs the Role engine," "Add a Flow constraint." Use the letters — I, C, R, A, F, T — as the class shorthand.
The 6-week iCRAFT lesson plan
One engine per week, in teaching order — from the most concrete (Instruction) to the most integrative (Transformation). Each week: lesson → guided practice → challenge → reflection.
| Week | Engine | Learning objective | In-class activity | Homework challenge |
|---|---|---|---|---|
| 1 | 🔵 Instruction | Write a command with a verb, deliverable, quantity, and audience. | "Fix the vague prompt" relay: teams upgrade 5 weak prompts against the clock. | Write 3 Instruction prompts for real schoolwork; score each in the Lab; screenshot best score. |
| 2 | 🟣 Creative | Add style, mood, and one creative constraint to a task. | Same topic, six moods: each group writes the same story prompt in a different tone; class compares AI outputs. | Design one image or video prompt using the 7-layer structure. |
| 3 | 🟠 Role-Based | Assign seniority + role + focus area to get expert-voice answers. | "Panel of experts": ask one question as 4 different roles; discuss how answers differ and why fact-checking matters. | Interview an AI "career coach" about a field you're curious about; note 2 claims to verify. |
| 4 | 🟢 Analytical | Demand evidence, structure, and visible reasoning. | Debate prep: two teams prompt for both sides of one question, using "present both sides fairly." | Build a comparison table prompt on a real decision (phone, laptop, college class). |
| 5 | 🔴 Flow/Steps | Break a goal into a capped, sequenced plan with checkpoints. | Reverse engineering: hand students a finished plan; they reconstruct the prompt that produced it. | Create a personal 5-step study plan prompt with time estimates and a success metric. |
| 6 | 🟩 Transformation | Convert content between formats while keeping facts accurate. | Telephone game: essay → thread → pitch → poem, verifying facts survive each hop. | Capstone: solve one real problem using at least 3 engines chained together; present the chain. |
Classroom prompt challenges
- Score Duel — two students, same task, 5 minutes. Highest Lab rubric score wins; loser explains what the winner did better.
- Minimum Viable Prompt — hit 80+ on the rubric in the fewest words possible.
- Engine Detective — show only an AI output; students identify which engine (and which ingredients) produced it.
- The Constraint Gauntlet — each round adds one constraint ("no adjectives," "exactly 5 bullets"); prompts must keep working.
- Chain Reaction — groups chain 3 engines (e.g., Analytical → Flow → Transformation) to go from question to finished deliverable.
AI literacy guidance to teach alongside
- Fact-check rules per engine — Instruction and Analytical outputs are verify-mandatory; Role-Based advice and statistics should be checked; Transformation must preserve core facts.
- A confident tone ≠ a correct answer — personas make style, not truth.
- Attribution & honesty — students disclose AI assistance and can explain their prompt chain.
- Privacy basics — never paste personal, private, or school-confidential information into a prompt.
- The human stays in charge — AI drafts; students decide, edit, and own the result.
The grading rubric (printable)
The same 7-criteria, 100-point rubric the Practice Lab uses — so students' self-practice and your grading speak the same language.
| Criterion | Pts | Full marks look like… | Zero looks like… |
|---|---|---|---|
| Action verb | 15 | Opens with a strong command verb. | A vague wish: "Can you maybe help with…" |
| Specificity | 15 | Exact deliverable + topic named. | "Write something about my project." |
| Role / voice | 15 | Seniority + role + focus area when judgment is needed. | No perspective on an expert-judgment task. |
| Constraints | 15 | Length cap, content rules, or creative limits. | No limits of any kind. |
| Output format | 15 | Shape declared: table, bullets, steps, email… | Format left to chance. |
| Context & audience | 15 | Who it's for, tone, and scenario. | No audience or tone anywhere. |
| Clarity & length | 10 | One readable instruction, ~15–60 words. | Two words, or a rambling wall of text. |
Suggested grade bands: 85–100 Excellent · 70–84 Strong · 50–69 Developing · below 50 Rebuild with the formula.
Printable student worksheet
iCRAFT Prompt Worksheet — Name: ______________ Engine: ______
- The question my engine asks: WHAT / WHAT+HOW / WHO / WHY+HOW / HOW / X→Y (circle one)
- My first-attempt prompt (before the lesson): ______________________________ Lab score: ____ /100
- The formula for my engine: ______________________________
- My upgraded prompt: ______________________________ Lab score: ____ /100
- What I added that raised my score: ______________________________
- One thing I must fact-check in the AI's output: ______________________________
Bring iCRAFT training to your school or program.
Dr. Beza runs live workshops and train-the-trainer sessions for educators and institutions.
Request classroom training