11 of 20 modules complete
This is the final module of Unit 3 — and the most directly actionable one. Over the past three modules you built prompts for developers, analysts, and managers. This module shows you how to collect, organise, and maintain all of them into a personal prompt library that compounds in value over time.
A prompt library is the single most practical career deliverable of this course. It is what separates someone who uses AI occasionally from someone who uses it systematically.
What is a prompt library and why does it matter?
A prompt library is a personal collection of tested, reusable prompts organised by task and role — your own reference system for getting consistent, high-quality AI output without writing prompts from scratch each time. Think of it as the AI equivalent of a code snippet library or a templates folder.
The value compounds over time. Your first 10 prompts save you 30 minutes a week. Your first 50 prompts save you 2–3 hours. Engineers and analysts who have been maintaining prompt libraries for a year report saving 8–10 hours a week — the equivalent of a part-time job.
There are three specific benefits beyond time saving:
Consistency — the same prompt produces the same quality output every time. You stop getting variable results because you stop writing variable prompts.
Institutional knowledge — your library captures what works for your specific context: your company, your clients, your tech stack, your role. Generic prompts from the internet do not have this. Yours do.
Career portability — your prompt library moves with you. When you change teams, companies, or roles, your library accelerates your ramp-up. It is a professional asset in the same way a portfolio is.
The prompt library entry format
Every prompt in your library should follow a consistent entry format. This makes it searchable, maintainable, and shareable with your team.
STANDARD PROMPT LIBRARY ENTRY FORMAT
## [Prompt Name — clear, specific]
Category: Developer / Analyst / Manager / Communication / Learning
Tool: ChatGPT / Claude / Gemini / Any
Last tested: [Month Year]
Use case: One sentence describing when to use this prompt
Prompt:
[The full prompt text, with [PLACEHOLDERS] for variable parts]
Notes:
- What this does well
- What to watch out for (common failure mode)
- Variations that work well
Example output quality: High / Medium (adjust with follow-up) / Low (starting point only)
The [PLACEHOLDERS] pattern is the most important element. Instead of a prompt that is hardcoded to one project, you write a reusable template:
Instead of: “You are a TCS delivery manager writing to Sarah Chen at UK Insurance about the API delay…”
Write: “You are a [ROLE] writing to [RECIPIENT] at [COMPANY] about [SITUATION]…”
This turns a one-time prompt into a reusable template for every similar situation.
Starter library — 10 prompts to begin with
Here are 10 fully written, ready-to-use prompts across the roles covered in Unit 3. Copy these directly into your library and customise the placeholders.
You are a senior [LANGUAGE] developer reviewing code for a [COMPANY TYPE] production system.
Review this code and respond in exactly this format:
- BUGS: Logic errors or runtime exceptions possible in production
- PERFORMANCE: Inefficiencies — N+1 queries, unnecessary loops, memory issues
- SECURITY: SQL injection, hardcoded secrets, unvalidated inputs
- READABILITY: Naming, structure, missing documentation
- VERDICT: Pass / Needs changes — one-line summary
Code:
[PASTE CODE HERE]
Claude preferred
Any language
Pre-commit review
I am debugging a [LANGUAGE/FRAMEWORK] issue.
Error: [PASTE EXACT ERROR OR TRACEBACK]
Code: [PASTE RELEVANT FUNCTION OR SECTION]
Context:
- Expected: [WHAT SHOULD HAPPEN]
- Actual: [WHAT IS HAPPENING]
- Already tried: [LIST ATTEMPTED FIXES]
- Stack: [LANGUAGE VERSION, FRAMEWORK, KEY LIBRARIES]
Identify the bug, explain why it occurs, and provide the corrected code.
Claude / ChatGPT
Any language
Save hours on debugging
You are a senior data analyst writing SQL for a [DATABASE TYPE] database.
Tables:
[PASTE YOUR SCHEMA — table names and column names]
Write a query that:
- [DESCRIBE WHAT YOU WANT IN PLAIN ENGLISH]
- [ADD FILTERS, DATE RANGES, AGGREGATIONS]
- Output format: [COLUMNS YOU WANT IN THE RESULT]
Include comments explaining each major section.
ChatGPT preferred
MySQL / PostgreSQL / BigQuery
Always verify output
You are a senior analyst writing for [AUDIENCE — e.g. "non-technical operations head"].
Here is the data: [PASTE TABLE OR KEY NUMBERS]
Write a [NUMBER]-paragraph summary.
- Each sentence must contain at least one number
- Tone: direct and factual, no corporate filler
- Paragraph 1: overall performance
- Paragraph 2: highlights and concerns
- Paragraph [LAST]: [NUMBER] specific recommendations
Under [WORD LIMIT] words.
Claude preferred
Reports / dashboards
Each sentence needs a number
You are a [MY ROLE] at [MY COMPANY] writing a professional email.
To: [RECIPIENT NAME, TITLE, COMPANY]
Situation: [DESCRIBE THE SITUATION IN 2–3 SENTENCES]
Key message: [WHAT YOU NEED THEM TO KNOW OR DO]
Tone: [PROFESSIONAL + SPECIFIC TONE NOTE]
Length: Under [WORD LIMIT] words
Do NOT use: “as per”, “please be advised”, “we sincerely apologise”, “going forward”
Include a clear subject line.
Claude preferred
Client / internal / vendor
Always add personal touch before sending
Convert these rough meeting notes into a professional summary.
Notes: [PASTE RAW NOTES]
Format:
- Meeting details (date, attendees, purpose — 1 line each)
- Decisions made (bullet list)
- Action items table: Owner | Action | Due date
- Issues / risks flagged
- Next steps
Keep language concise. Flag any ambiguous items with [NEEDS CLARIFICATION].
Claude / ChatGPT
Any meeting type
[NEEDS CLARIFICATION] flag is key
You are a project manager writing a weekly status report.
Project: [PROJECT NAME AND BRIEF DESCRIPTION]
Audience: [CLIENT / MANAGEMENT — technical level]
Week ending: [DATE]
Overall status: [Red / Amber / Green — and why it changed if it did]
Input data:
- Milestones: [STATUS, % COMPLETE, PLANNED VS FORECAST DATE]
- Issues: [LIST WITH SEVERITY]
- Risks: [LIST WITH MITIGATION STATUS]
- Highlights: [THIS WEEK’S WINS]
- Next week focus: [TOP 3 PRIORITIES]
Generate report with: Executive summary (3 sentences), milestone table, issues and risks (RAG labels as text), highlights, next week focus.
Use RAG as text labels (Red/Amber/Green) — not HTML colours.
Under 400 words total.
Claude preferred
Indian IT delivery
RAG as text for Outlook compatibility
You are an experienced [DOMAIN] professor explaining a concept to a final-year engineering student preparing for a campus placement interview.
Explain: [CONCEPT]
Structure:
- One-line definition (plain language, no jargon)
- Real-world analogy relevant to Indian context
- How it works — step by step (maximum 4 steps)
- One concrete code or application example
- Two likely interview questions on this topic with model answers
- Common mistake candidates make when answering
Keep total length under 500 words.
Any tool
Technical concepts
Interview prep powerhouse
Write a LinkedIn post about [TOPIC / ACHIEVEMENT / LEARNING].
My context: [1–2 sentences about who you are and why this matters to you]
Style requirements:
- Opens with a hook (surprising fact, counterintuitive statement, or short story)
- 150–200 words
- Short paragraphs (2–3 sentences max each)
- Ends with a question or call to action
- Sounds personal, not corporate
- No hashtag spam (maximum 3 relevant hashtags at the end)
Do NOT start with “I am excited to share” or “I am thrilled to announce”.
Claude preferred
Always edit to add your voice
Test: would you actually post this?
You are a rigorous devil's advocate. Your job is to find every weakness in what I am about to share.
[PASTE YOUR PLAN, PROPOSAL, ARGUMENT, OR DECISION]
Identify:
- The 3 strongest objections a sceptic would raise
- Assumptions I am making that could be wrong
- Edge cases or scenarios where this fails
- What I would need to prove to make this airtight
Do NOT agree with me. Do NOT soften your critique. Be specific — general criticism is not useful.
Claude preferred
Plans / proposals / decisions
Use before any important submission
Where to store your library
Notion
Best overall option. Create a database with tags for Category, Tool, and Quality rating. Full-text search across all prompts. Accessible on mobile. Can share with teammates.
RecommendedGoogle Docs
Simple and accessible anywhere. One doc per category (Developer, Analyst, Manager). Use headings for Ctrl+F search. Easiest to start with if you already use Google Workspace.
Good startObsidian / Markdown
Best for technical users. Local storage, full control, works offline. Tag system for filtering. Can sync via iCloud or Dropbox. Ideal if you already use Obsidian for notes.
For power usersAvoid storing prompts in chat history. Chat history is not searchable, not organised, and gets lost. Your library needs to be a document you intentionally maintain.
Prompts by role — what your library should contain
Use this checklist to identify gaps in your library. Aim to have at least one tested prompt per category before Unit 5.
💻 Developer prompts
- Code generation (your stack)
- Universal debug template
- Code review (pre-commit)
- Unit test generation
- Docstring / README generation
- SQL generation (your DB)
📈 Analyst prompts
- SQL from schema (your tables)
- Data insight summary
- Anomaly identification
- Python analysis boilerplate
- Report narrative template
- Slide content from findings
📋 Manager / communication prompts
- Difficult email template
- Client status update
- Meeting notes → summary
- Weekly status report
- Escalation email
- QBR / presentation outline
🎓 Learning / career prompts
- Concept explainer for interviews
- Mock interview question generator
- LinkedIn post (your style)
- Resume bullet rewriter
- Devil's advocate for plans
- Company AI research template
How to grow and maintain your library
A prompt library that is not maintained quickly becomes a graveyard of prompts you do not trust. Three habits keep it alive:
1. Add immediately, refine later. When a prompt works well — even once — add it to the library immediately with a note on what it did well. You can refine the wording later. The barrier to entry should be as low as possible.
2. Rate every prompt after testing. Use a simple quality rating: High (use as-is), Medium (use with one follow-up), Low (starting point only). Prompts that consistently land at Low should be rewritten or deleted.
3. Review and update quarterly. AI tools change fast. A prompt that worked best in Claude Sonnet 3.5 may work better in ChatGPT after a model update. Review your library every three months, test your top 10 prompts, and update the “Last tested” date and tool recommendation.
Unit 3 — What you have learned
MODULE 3.1
GenAI for Developers
Code generation, debugging, code review, test writing, documentation. Read every line before committing. Specificity produces production-ready output.
MODULE 3.2
GenAI for Analysts
SQL from schema, data summarisation, Python analysis, report narratives. AI writes the draft — you provide the business judgement. Verify every query.
MODULE 3.3
GenAI for Managers
Professional emails, meeting summaries, status reports, presentations. Context and relationship define quality. Always add one personal touch before sending.
MODULE 3.4
Your Prompt Library
A personal, organised, maintained collection of tested prompts. The most compounding career asset from this course. Starts today, grows with every task.
Unit 4 goes deeper — into RAG, AI agents, and automation workflows. The prompt library you build today will be extended with advanced technique prompts after Unit 4.
The three things to remember
KEY TAKEAWAYS — MODULE 3.4
- Use placeholders, not hardcoded prompts — a prompt with [ROLE], [COMPANY], and [SITUATION] placeholders is 10x more useful than one written for a single occasion. The 30 seconds it takes to generalise a prompt pays back every time you reuse it.
- Store it where you will actually find it — a prompt library in a document you open daily is worth more than a perfect system you never use. Start with Google Docs if that is what you already use. Migrate to Notion when you outgrow it.
- Your library compounds — 10 prompts save 30 minutes a week. 50 prompts save 2–3 hours. 100 prompts save half a workday. The professionals getting the most value from AI in 2026 are not using better tools — they are using the same tools with a personal system built over months of deliberate practice.
This is the Unit 3 capstone assignment. It consolidates everything from Modules 3.1, 3.2, and 3.3 into a single deliverable you will actually use.
- Choose your storage tool — Notion, Google Docs, or Obsidian. Create a new document or database called "My Prompt Library." Set up the standard entry format from this module.
- Add the 10 starter prompts from this module — customise every [PLACEHOLDER] for your specific context: your role, your tech stack, your target company, your domain. A prompt that still has generic placeholders is not in your library — it is a template.
- Add your best prompts from Modules 3.1, 3.2, and 3.3 assignments — each assignment asked you to save prompts in your AI Learning Journal. Transfer the best ones to your library using the standard format. Test each one and rate it High / Medium / Low.
- Identify 3 gaps — using the role checklist in this module, identify 3 prompt categories you need but do not yet have. Write those prompts from scratch, test them, and add them to the library.
- Target: 15 prompts minimum by the end of Unit 3. By Unit 5, you will extend this to 25+ prompts covering your full workflow.
Your prompt library is a living document. The most important habit is adding to it immediately when something works — before the next meeting, the next task, the next distraction takes over. Start that habit today with this assignment.
Sources and references
| Claim / Component | Source / Note |
|---|
| Prompt library value compounding — 10 prompts → 50 prompts → time savings | Synthesised from Lakera Prompt Engineering Guide 2026; promptbuilder.cc Best Practices 2026 |
| Notion recommended for prompt libraries; Obsidian for power users | Community consensus from prompt engineering practitioner communities 2025–2026 |
| 10 starter prompts — all original, tested against real Indian IT work contexts | Original content — CC0, no attribution required |