Edor.ai Learning Centre
Prompt Engineering Classroom: what it looks like in a lesson
Eighteen lessons, all set in Hong Kong classrooms: planning, tiered worksheets, question setting and rubrics, marking feedback, SEN adaptation, guided practice, administrative writing, grounded RAG answers, defending against prompt injection, and keeping a panel template library alive. Every lesson comes with prompts you can copy, in both languages.
18 lessons
about 241 minutes
For: classroom teachers, panel heads, teaching assistants; read LLM Classroom modules 1 and 4 first
Syllabus
01
Module 1: The fundamentals
Frameworks, worked examples, chain of thought and system instructions
- 01Where a good prompt startsA Primary 4 Chinese Language oral-assessment example shows the gap between a one-line request and a prompt that states audience, purpose, class profile and return format — plus a map of the four course modules. 12 min· Level: Foundation
- 02The lesson-planning work orderReplace vague instructions with a five-field work order — recipient setup, procedure, field intel, delivery spec and red lines — demonstrated on Primary 5 English reading, Secondary 2 Science labs and Secondary 4 Citizenship. 14 min· Level: Foundation
- 03Lock the format with samplesWhen the spec has been rewritten three times and the output still drifts, you probably need demonstrations, not adjectives. Three scenarios — Secondary 2 Geography map reading, Primary 6 Maths working layout and English feedback register — show how two or three examples plus a counter-example lock item type, tone and layout. 13 min· Level: Hands-on
- 04Steps before the answerWrong word-problem answers and long tasks that fall apart halfway usually mean the model was allowed to conclude too early. Primary 5 fractions, a marker review pass and a Primary 4 STEM project show step-first output, role switching and task decomposition. 14 min· Level: Intermediate
- 05System prompts and school personasA system prompt is a standing instruction that applies to every conversation. Write your school context, language, terminology and safety floor into it once, and teachers only ever type the day's task. 13 min· Level: Intermediate
02
Module 2: The situations you are actually in
Planning, question setting, marking, tiering, guided practice, and what students themselves need to understand
- 06Lesson planning promptsPlanning is slow because you start from blank every time, not because you type slowly. Four templates — the plan, a slide outline, upgrading an old plan, and post-lesson reflection — to use alongside our 30-minute workflow. 14 min· Level: Hands-on
- 07Worksheet and differentiation promptsThree-tier worksheets fail when the only difference is question count. Here is how to tier by cognitive level and support instead, with templates for language support and SEN adaptation. 14 min· Level: Hands-on
- 08Question setting and rubric promptsA good paper is derived from learning focuses and cognitive levels, not from whatever comes to mind. Four prompts — specification table, items from the table, rubric, and paper review. 15 min· Level: Hands-on
- 09Marking and feedback promptsPosition AI as a provisional marking assistant for teacher review, with prompts for Chinese expository writing, English marking codes, mathematical working diagnosis and whole-class error patterns, plus the baseline for de-identifying student work. 16 min· Level: Hands-on
- 10Scaffolded tutor prompts that withhold answersA complete tutor system prompt that turns a general AI from an answer machine into a scaffolded tutor working up four hint levels, with a worked Secondary 1 science transcript and how the teacher sets the hint ceiling. 14 min· Level: Hands-on
- 11Teaching students prompt literacyA classroom design you can run as written, with a student worksheet prompt, a method for checking AI claims against original sources, and a specific AI-use declaration template plus an integrity discussion frame. 13 min· Level: Hands-on
03
Module 3: Paperwork, RAG and safety
School documents, grounded answers, prompt injection, and content gating on the student side
- 12School administration and parent communication promptsWriting administrative prompts with placeholders so identifiable student and staff data never reaches a general-purpose tool, with ready templates for circulars, letters to parents, minutes and a parents' evening script. 12 min· Level: Hands-on
- 13Grounded answers from school documents onlyConfine the model to the passages you supply, require a source number and a verbatim quote for every claim, and authorise it to answer "not found in the provided documents" — the most effective anti-hallucination pattern for school policy and admin queries. 14 min· Level: Intermediate
- 14Prompt injection and student content safetyHomework, compositions and external web pages can carry sentences that try to change how the model behaves. This lesson covers delimiters, instruction hierarchy, and reporting instead of obeying — and is honest that prompting alone is not enough. 14 min· Level: Intermediate
04
Module 4: Going further, and keeping it working
Structured output, multimodal prompts, running a template library, and a debugging checklist
- 15Structured output when thirty runs must mergeFrom markdown tables to strict JSON — how to specify column names, order and permitted values, why drifting columns hurt more than wrong wording, and what to do when commentary wraps the JSON. 12 min· Level: Intermediate
- 16Multimodal prompts: transcribe first, judge secondHandwritten work, graphs and reading recordings need a split between what the model saw or heard and how it marks or feeds back — with honest limits on Chinese handwriting and accented speech. 12 min· Level: Hands-on
- 17Build a panel template library the school keepsFour-part names, complete metadata and a termly panel review turn strong prompts in private chats into school assets that survive staff turnover. 10 min· Level: Hands-on
- 18Debug prompts with six ordered checksWhen output fails, do not rewrite the whole prompt. Work through task clarity, format, context length, buried instructions, example conflicts and model fit — one change at a time. 15 min· Level: Intermediate