Hand-coded HTML · Performance task
Prompt Lab
A hands-on lab where staff build prompts piece by piece, watch the AI reply improve, then prove the skill in three rubric-scored tasks.
The challenge
Staff had access to an AI assistant but got generic, overlong or made-up answers, and blamed the tool. The gap was prompting skill, and a multiple-choice quiz about prompting would not show whether people could actually do it.
My approach
- A builder where learners switch on role, context and format, and watch a scripted reply and quality meter change, so cause and effect is visible.
- Colour-coded prompt parts carried through every screen, so the structure becomes a mental model.
- Three performance tasks against a rubric shown up front: rebuild weak prompts, fix flawed outputs, and match tasks to patterns.
- An ungraded prompt checker for learners' own prompts, and a printable cheat sheet.
Objectives and how each is assessed
Every assessment item was written before the content, and maps to exactly one objective.
| Learning objective | Assessment |
|---|---|
| Structure a prompt with role, context, task and format. | Rebuild 3 weak prompts, choosing the strongest option for each part (12 points) |
| Refine a weak AI output with a targeted follow-up prompt. | Choose the best follow-up for 3 flawed outputs (6 points) |
| Choose the right prompt pattern for summarising, drafting, analysing and brainstorming. | Match 4 work tasks to patterns (4 points); pass mark 18 of 22 |
Results and reflection
Add pilot results, reviewer quotes or what you would do differently next time.
Design artefacts
- Prompt cheat sheet (printable)
- Storyboard (available on request)