A strong AI prompt states the task, gives context (who you are and why), supplies the input in clear delimiters, says what the output should look like and sets rules such as "use only the text below". Below are 18 prompts for study, coding, debugging, resumes, research, data, interviews and writing, each shown next to the weak version, plus six templates you can copy.
Replace the [brackets] with your own details and remove personal data before you paste. Several prompts ask for feedback or practice questions instead of finished answers, on purpose: that keeps the work yours and is safer for academic integrity. Read what prompt engineering is first if you want the reasoning behind them.
Study and revision
Study: revise one exam topic (Operating Systems, deadlock)
Weak: Explain deadlock
I'm a 2nd-year B.Tech CS student revising Operating Systems for a university exam next week. Explain deadlock in under 200 words: definition, the four necessary conditions, and one everyday analogy. Then ask me 3 short-answer questions on it, without answers. After I reply, mark my answers and tell me what I missed.Why it works: States level, goal, length and output parts (clear and direct), and turns the chat into practice you do yourself instead of passive reading.
Study: build a revision plan (DBMS, 10 days)
Weak: Make me a study plan for DBMS
I have 10 days before my DBMS exam: 2 hours on weekdays, 4 on weekends. Topics: ER model, normalization, SQL, transactions, indexing. I'm weak at normalization and transactions. Make one table with day, topic, task and a 15-minute self-test. Put weak topics early and keep day 10 for revision only.Why it works: Gives real constraints (time, weak areas) and a fixed output format, so the plan is usable instead of generic.
Study: quiz yourself from your own class notes
Weak: Make a quiz from this
My class notes are inside <notes> tags. Write 8 MCQs with 4 options each, using only these notes. After all questions, give the answers with a one-line reason pointing to the note it came from. If the notes don't cover something, don't invent a question for it.
<notes>[paste notes]</notes>Why it works: Delimiters separate your notes from instructions; 'only these notes' grounds the answers and allows 'not covered'.
Coding, debugging and learning a library
Coding: write a function from a clear spec
Weak: Write a function to validate email in python
Write a Python 3.12 function validate_email(s: str) -> bool for a signup form. Rules: exactly one @, non-empty local part, domain contains a dot, no spaces. Use only the standard library. Add type hints, a docstring and 6 pytest cases (3 valid, 3 invalid, one an edge case). Then list any rule you had to assume.Why it works: A precise spec plus tests gives you something checkable, and asking for assumptions exposes gaps early (clear and direct, self-check).
Coding: get a useful code review of your own project
Weak: Is my code good?
Review this Python function from my college project. Check in this order: correctness bugs, edge cases (empty list, None), readability. For each issue give the line, the problem and a 1-2 line fix. Don't rewrite the whole function. If you find no bugs, say so instead of inventing problems.
<code>[paste]</code>Why it works: Ordered criteria and a 'say so if none' escape hatch reduce padded, made-up feedback.
Debugging: an error you cannot read
Weak: My code isn't working, fix it
My Flask app returns a 500 error on POST /login (Python 3.11, Flask 3.0). Traceback and code are below. First list the 3 most likely causes, ranked, with the line of the traceback that supports each. Then give the smallest fix for the top cause and tell me how to confirm it worked.
<traceback>[paste]</traceback>
<code>[paste]</code>Why it works: Environment, exact error and a ranked-hypotheses request beat 'fix it'; you also learn how to verify the fix.
Debugging: wrong output, no error
Weak: Why is my loop giving the wrong answer
This function should return the second-largest number in a list, but for [3, 9, 9, 5] it returns 9; I expect 5. Explain in plain language why it fails, give corrected code, and add 3 test inputs that would have caught this bug, including duplicates and a one-element list.
<code>[paste]</code>Why it works: Expected vs actual output pins the bug down; the test request makes the model prove the fix and teaches you to write tests.
Learning a new library (pandas) in one afternoon
Weak: Teach me pandas
I know basic Python but have never used pandas. I have 3 hours. Give a 5-step path built around one real task: cleaning a CSV of student marks (missing values, duplicate rows, wrong data types). For each step: one concept, 5-10 lines of pandas 2.x code, and one exercise for me. Don't give exercise solutions until I ask.Why it works: Level, time box, one concrete task and library version keep it focused; withholding solutions forces you to practise.
Resume and interview preparation
Resume: rewrite bullets without inventing facts
Weak: Improve my resume
Rewrite the 4 bullets below for a data analyst internship: action verb, what I did, result, under 20 words each. Use only facts I've given; where a number is missing put [ADD NUMBER] instead of guessing. Then ask me 3 questions that would help me fill those gaps.
<bullets>[paste]</bullets>Why it works: Truthfulness rule plus placeholders stop fabricated metrics, which interviewers will probe.
Resume: tailor to one job description honestly
Weak: Make my resume match this job
Compare my resume and the job description below. 1) List 5 skills the JD stresses that my resume already shows, quoting the line. 2) List 3 it doesn't show. 3) For each gap, say where I could add it only if I truly have that experience; otherwise suggest a small project that would build it. Don't add skills I haven't listed.
<resume>[paste]</resume>
<jd>[paste]</jd>Why it works: Splits the task into checkable steps, grounds claims in quoted lines, and blocks keyword-stuffing.
Interview prep: a mock technical interview
Weak: Ask me interview questions
Run a mock interview for a Junior Backend Developer role (Node.js, SQL) at a product startup. Ask one question at a time and wait for my answer. After each answer give a score out of 5, one thing I did well, one gap and a stronger model answer in 3 sentences. Start easy and get harder over 6 questions. Begin now.Why it works: Defines the loop (one question, wait, feedback format), which a one-line request cannot.
Interview prep: 'tell me about yourself' from your real facts
Weak: How do I answer tell me about yourself
Draft a 60-second 'tell me about yourself' for campus placements. My facts: final-year B.Tech CSE, built a library app in Django, 2 hackathons, 2-month QA internship at a local startup. First person, spoken style, max 130 words, ending with why I want a software role. Don't add facts. Then give 2 follow-up questions an interviewer might ask.Why it works: Supplies true raw material and limits, so the output is yours to rehearse, not a generic script.
Research and data analysis
Research: starting a literature search for a project
Weak: Give me research papers on federated learning
I'm starting a final-year project on federated learning in healthcare. Suggest 5 well-known survey or foundational papers (title, authors, year) and what each is good for. Include only papers you are confident exist; mark any doubtful one 'verify'. I will check every title on Google Scholar. Add 4 search keywords for finding newer work.Why it works: Permission to flag uncertainty cuts invented citations; you still verify each paper yourself.
Research: summarise a paper you pasted
Weak: Summarize this paper
Summarize the paper below for a classmate who knows basic ML. Give: the problem (1 sentence), the method (2-3 sentences), the main result with the exact numbers, and 2 limitations the authors state. Quote the sentence that supports each number. If the paper doesn't say something, write 'not stated'.
<paper>[paste]</paper>Why it works: Quote-first grounding plus 'not stated' makes the summary auditable against the source.
Data analysis: plan before you code (placement data)
Weak: Analyze my data
Below are the column names and 5 sample rows of a college placement dataset (no personal data). Goal: find what is associated with getting placed. Suggest a step-by-step plan in pandas: cleaning checks, 3 charts, 1 simple model, and how I'll check I'm not misled (class imbalance, small sample). Write code for step 1 only.
<data>[paste]</data>Why it works: States the goal and data shape, asks for a plan with self-checks, and uses one step at a time (decomposition).
Data analysis: a SQL query you can verify
Weak: Write SQL for top students
Tables: students(id, name, dept), marks(student_id, subject, score). Write a PostgreSQL query for the top 3 students per department by average score, ties broken by name. Explain each clause in one line and show a 2-row sample output so I can check it. If something is ambiguous (e.g. students with no marks), state your assumption.Why it works: Schema, dialect and tie rule remove guesswork; sample output and stated assumptions let you test the answer.
Writing feedback
Writing feedback: improve your statement of purpose paragraph
Weak: Fix my essay
Give feedback on my SOP paragraph below; don't rewrite it. Tell me: (1) its main point in one sentence, so I can check it's clear; (2) the weakest sentence and why; (3) two places that need a specific detail; (4) any sentences that sound generic. Keep my voice. End with 2 questions I should answer in my revision.
<draft>[paste]</draft>Why it works: Asks for diagnosis, not a rewrite, so the work and the learning stay yours (also safer for academic integrity).
Writing feedback: a short email to a professor
Weak: Write email to professor for internship
Write a short email to Prof. [Name] asking about a research internship. Facts: I'm a 3rd-year ECE student at [College]; I read their 2024 paper on [topic]; I built [project]. Under 120 words, polite, one clear ask: a 15-minute call. Include a subject line. Don't flatter or exaggerate; leave [brackets] for anything I haven't given.Why it works: Real facts, a single ask and a word limit; the bracket rule stops invented details.
Six reusable templates
Each template can be pasted into any chatbot. Replace the brackets.
Task - Context - Input - Format - Rules (everyday default)
Task: [what you want done, in one sentence]
Context: [who I am, why I need it, who will read it]
Input:
<input>[paste text, code or data]</input>
Output format: [bullets / table / 150 words / code only]
Rules: [length, tone, things to avoid, 'use only the input']
If something is missing or unclear, ask me before answering.Few-shot (show, don't describe)
Task: [what to do with each input].
<examples>
<example><input>[example 1]</input><output>[ideal output 1]</output></example>
<example><input>[example 2, different kind]</input><output>[ideal output 2]</output></example>
<example><input>[example 3, edge case]</input><output>[ideal output 3]</output></example>
</examples>
Now do the same for:
<input>[your real input]</input>Grounded answer with quotes (for notes, papers, policies)
Answer the question using only the document below.
<document>[paste]</document>
Question: [your question]
First quote the passages that matter, then answer and refer to the quotes. If the document doesn't contain the answer, say 'not in the document'. Do not use outside knowledge.Draft - review - refine (self-check chain)
Step 1 (separate message): Write a first draft of [item] for [audience], max [N] words.
Step 2: Review your draft against these criteria: [list 3-5 checks, e.g. accurate, specific, no jargon, fits the word limit]. List the problems.
Step 3: Rewrite the draft fixing those problems. Give me the final version and a 3-bullet list of what changed.Socratic tutor (learn by answering)
Teach me [topic] at [level, e.g. first-year B.Tech]. Ask me one question at a time and wait for my answer. After each answer: say what is right, correct what is wrong in 2 sentences, then ask a slightly harder question. Don't reveal answers before I try. After 6 questions, list my weak areas and what to revise.Structured extraction (JSON or table)
Extract the fields below from the text.
<text>[paste]</text>
Return only valid JSON matching this shape: {"name": string, "date": "YYYY-MM-DD" or null, "amount_inr": number or null, "summary": string (max 20 words)}
Use null when a field is not in the text. Don't guess. I will validate the output before using it.How to adapt any prompt
- Replace the details with your own level, subject, tool and deadline.
- Keep the output format line, because it makes the answer checkable.
- Add "if something is missing, ask me first" when the task is complicated.
- Run it twice. If the answers differ a lot, tighten the constraints.
- Check facts, code and numbers yourself before you rely on them, and follow your institution's rules on AI use.
Frequently asked questions
What is an example of a good AI prompt?
"I'm a 2nd-year B.Tech student revising Operating Systems. Explain deadlock in under 200 words: definition, the four necessary conditions and an everyday analogy. Then ask me three short-answer questions without answers." It names the level, goal, length and output parts.
How do I write a prompt for coding?
Give the language and version, the exact behaviour you want, constraints, and ask for tests and any assumptions. Paste the error message and code in delimiters when debugging.
Can I use these prompts with ChatGPT, Gemini and Claude?
Yes. They follow guidance common to the main vendors. Models differ in style, so adjust the length and format lines if a tool answers too briefly or too much.
Is it okay to use AI prompts for assignments?
Follow your institution's policy and disclose AI help where required. Prompts that ask for feedback, hints or practice questions keep your own work central.
Why do my prompts give generic answers?
They usually lack context, an audience or a format. Add who you are, what you need it for and the exact shape of the answer.
Learn this with mentors
AI and machine learning programmes with admissions open: live classes, mentors and project work.
