Prompt Engineering Masterclass

0) How to Use This Material

Use this as self-study, a 2–4 hour facilitated workshop, or a team training pack.

Prerequisites: none — suitable for students, professionals, and beginners.

1) Learning Outcomes

1. Explain prompt engineering and why it matters.
2. Diagnose weak prompts and improve them systematically.
3. Use multiple prompting patterns (few-shot, rubric-driven, iterative).
4. Build prompts for real tasks across domains.
5. Evaluate AI outputs using a practical rubric.
6. Run an end-to-end prompt workflow: define → craft → evaluate → iterate.

2) What Prompt Engineering Really Is

Prompt engineering designs inputs that reliably produce useful AI outputs. It's about clear goals, constraints, context, and evaluation criteria — not secret words.

Mental Model

• Treat the AI like a capable but literal collaborator.
• It can reason with what you provide but won't infer missing requirements.
• Explicit structure and requirements yield best results.

Why This Skill Matters

3) Prompt Anatomy: RTFCC

Base Template

You are a [role].
Task: [clear objective].
Context:
- [fact 1]
- [fact 2]
- [fact 3]
Constraints:
- [tone/length/reading level]
- [must include]
- [must avoid]
Output contract:
- [format]
- [sections]
- [quality criteria]
If needed, ask [N] clarifying questions before answering.

4) Levels of Prompt Quality

Level 1 — Request

Write about remote work.

Problem: vague task, no audience, no structure.

Level 3 — Structured Prompt

You are an HR advisor.
Explain remote work to new managers.
Include 4 benefits, 4 challenges, and 5 practical policies.
Keep it under 350 words in simple language.
Output as headings + bullet points.

Level 4 — Production-Ready

You are an HR advisor with policy-writing experience.
Task: Draft a manager guide for hybrid work onboarding.
Context:
- Company size: 80 employees
- Teams: engineering, sales, support
- Current issue: inconsistent work-from-home expectations
Constraints:
- Audience: first-time managers
- Tone: practical and supportive
- Length: 450-550 words
- Avoid legal claims and country-specific legal advice
Output contract:
- Section 1: Policy principles (5 bullets)
- Section 2: Team agreement template (checklist)
- Section 3: First 30-day rollout plan (week-by-week)
- Section 4: 3 risks and mitigations
Before drafting, ask 3 clarifying questions about local policies and team schedules.

5) Prompting Patterns

Direct Instruction

Generate 10 customer service email subject lines for delayed delivery notices.
Tone: empathetic and professional.
Max 8 words each.

Role + Audience Alignment

You are a cybersecurity trainer.
Explain phishing to office staff with no technical background.
Use everyday examples and a 5-item safety checklist.

Few-Shot Prompting

Provide input/output examples, then ask model to transform.
Useful for style matching and consistent tone.

Critique-and-Revise

Draft, self-critique for clarity and actionability, then revise.

Rubric-Guided Generation

Provide a rubric and require the model to score its own output and revise until scores meet thresholds.

6) Advanced Principles

• Specify the decision context (where the output will be used).
• Separate facts from instructions to avoid confusion.
• Constrain scope by geography, audience, timeframe, budget.
• Request verification behavior: ask assumptions and mark uncertainty.
• Force structured output when downstream reuse is required (JSON/tables).

7) Real-World Use Cases (Selected)

Learning

You are a learning strategist and tutor. Teach opportunity cost in plain language with analogies, a 5-question quiz, and a practical exercise.

Professional Email

You are a communications advisor. Draft an email to a vendor who missed two milestones. Tone: firm but respectful. Length: 140-180 words.

Small Business Plan

You are a small business advisor. Build a 30-day launch plan for a home-based fresh juice service. Budget: $400. Capacity: one person.

8) Prompt Evaluation Rubric

Score outputs 1–5 on the following criteria and sum:

1. Task Alignment
2. Accuracy
3. Completeness
4. Clarity
5. Actionability
6. Format Compliance

Interpretation: 26–30 production-ready; 20–25 useful with edits.

9) Common Failure Modes & Fixes

• Generic Output: add audience, goal, constraints.
• Wrong Tone: specify tone and do/don't examples.
• Missing Sections: add an explicit output contract.
• Hallucinated Facts: require assumptions and flag uncertainty.
• Verbose Output: add length bounds and structure.

10) Iterative Prompt Workflow

1. Define success in one sentence.
2. Draft Prompt v1 using RTFCC.
3. Run and evaluate with the rubric.
4. Diagnose gaps (task/context/constraints/format).
5. Revise 1–2 variables to create Prompt v2.
6. Save the final prompt as a reusable template.

11) Teaching Plan (Instructor)

1. Concept Lecture — 35 min
2. Live Demonstration — 20 min
3. Guided Practice — 35 min
4. Break — 10 min
5. Use-Case Labs — 35 min
6. Capstone Challenge — 35 min
7. Debrief — 10 min

Materials: this module, an AI assistant tool, rubric sheet, and peer feedback form.

12) Mini Tasks (Drills)

• Prompt Diagnosis — identify 5 missing elements; rewrite.
• Audience Shift — write two versions for children and executives.
• Format Control — produce checklist, 7-day plan, FAQ.
• Clarification Gate — force AI to ask questions first.
• Reliability Guardrails — add assumptions and "don't invent data".

13) Scenario Exercises

Student Support

Create a 14-day study plan (2 hours/day), stress strategy, and tracker.

Team Productivity

Create a meeting agenda template, decision log, and action tracker.

Freelance Growth

Build a 60-day client growth plan under $150 budget.

14) Capstone Project

Create a production-ready prompt system with: problem statement, Prompt v1, sample output, rubric scoring, Prompt v2, and a short reflection.

Evaluation: problem clarity, prompt quality, output usefulness, iteration quality, reflection depth, and format discipline.

15) One-Page Prompt Cheat Sheet

[ROLE]
You are a [specific role].

[TASK]
Do [specific objective] for [specific audience].

[CONTEXT]
- Situation:
- Inputs/data:
- Constraints (time, budget, tools):

[CONSTRAINTS]
- Tone/style:
- Length:
- Must include:
- Must avoid:

[OUTPUT CONTRACT]
- Format:
- Required sections:
- Quality criteria:

[SAFETY/RELIABILITY]
- State assumptions.
- Mark uncertainty.
- Do not fabricate specific data.

[INTERACTION RULE]
Ask [N] clarifying questions first when requirements are unclear.

16) Final Notes

Prompt engineering is applied thinking: define success, structure instructions, evaluate, and iterate. Complete the drills and capstone to gain transferable skill.

Interactive Quiz (Mixed Question Types)

Objective questions are auto-graded. Open-ended questions include a self-check rubric.