Two students open ChatGPT on the same night before the same exam. The first types, “Give me notes on computer networks.” The second types, “I am a Class 11 NEB student. Make short revision notes on computer networks with one example per topic, in simple English, and end with 5 MCQs.”
The second student gets something usable. The first gets a generic wall of text that may not match the NEB syllabus at all. Same tool, same minute, very different result. The only difference is the prompt.
This guide shows you how to write the second kind of prompt, even if you have never used AI before. We will use four small frameworks (TAG, CARE, RACE, and RISE) and apply them to three things Nepali students and job seekers actually need: exam prep, study notes, and resumes.
What a prompt is?
What is a prompt?
A prompt is the message you type to an AI tool. That is all. But the AI cannot see your classroom, your syllabus, your exam date, or your job application. It only knows what you tell it. If you leave things out, it fills the gaps with the most average guess possible.
So good prompting is not about secret tricks. It is about telling the AI the things you would tell a good tutor before asking for help: who you are, what you need, what it should look like, and how much time you have.
The four frameworks at a glance
| Framework | What the letters mean | Use it when |
|---|---|---|
| TAG | Task, Action, Goal | You need a quick, focused answer |
| CARE | Context, Action, Result, Example | The format of the answer matters |
| RACE | Role, Action, Context, Expectation | You want the AI to think like a specific expert |
| RISE | Role, Input, Steps, Expectation | You already have something and want it checked or improved |
These frameworks are popular in the prompt-writing community, and a clear graphic of them was shared by Analytics Vidhya. Below, each one gets a Nepal-based example you can copy and change.
You do not need to memorize the letters. They are just a checklist so you do not forget a part.
TAG: for quick, focused answers
TAG is the shortest framework. You say what the task is, what action you want, and what the goal is.
- Task: what you are working on
- Action: what the AI should do
- Goal: what you want to walk away with
Example: last-minute exam revision
Prepare me for my NEB Class 12 Physics exam in 3 days.
Summarize the Current Electricity chapter and point out the numerical types that come most often.
Give me 3 priority topics, 2 common mistakes students make, and 1 practice numerical with a solution.Notice the last line. “3 priority topics, 2 common mistakes, 1 practice numerical” gives the AI a clear shape to fill. Asking for numbers keeps the answer short, which is exactly what you want the night before an exam.
TAG works well for: revision plans, quick explanations, a one-page summary, or “what should I study first?”
CARE: when the shape of the answer matters
CARE adds context and an example of what you want. It is the best framework for study notes, because notes are only useful if they are in a format you can actually revise from.
- Context: who you are and the situation
- Action: what to do
- Result: what the finished output should look like
- Example: a style or sample to follow
Example: study notes
I am a Class 11 NEB Computer Science student in Nepal preparing for the final exam.
Summarize the main topics of the Computer Networking chapter.
I want a one-page revision sheet that covers network types, topologies and the OSI model, with the difference between each pair of similar terms.
Use bullet points, keep sentences short, and include one real example for each topic, like a school computer lab or a cyber cafe.The “example” part is where you can steer the style. You can also paste a sample of how you like your notes and say, “Follow this format.“
A tip that helps a lot: tell the AI the language. You can write, “Explain in simple English and add the Nepali term in brackets for difficult words.” Many students find this makes technical chapters much easier.
RACE: when you want an expert’s point of view
RACE starts with a role. Giving the AI a role changes how it answers, because it picks the vocabulary and priorities of that role.
- Role: who the AI should act as
- Action: what to produce
- Context: the details of your situation
- Expectation: tone, length, and format
Example: Loksewa or TSC exam practice
Act as an experienced Loksewa exam trainer for computer operator posts.
Write 10 multiple-choice questions on MS Word and MS Excel.
The student is a beginner who has about 2 weeks left and finds shortcut keys hard to remember.
Give the answer key with a one-line explanation for each answer. Keep the questions at the level of a real first-paper exam.You can also use RACE for interview practice. Ask the AI to act as an interviewer at a Kathmandu IT company and ask you one question at a time. Answer each one, then ask it to rate your answers. It is free practice, and nobody is watching.
RISE: when you already have something to improve
RISE is for checking and fixing your own work. You give the AI an input, ask for specific steps, and describe the output you want.
- Role: the expert who will review it
- Input: the thing you are giving it
- Steps: what to do, in order
- Expectation: how to present the result
Example: improving a resume
Most fresh graduates in Nepal write a resume that lists duties instead of results. “Responsible for managing the college website” tells an employer very little. RISE helps you find and fix lines like that.
Act as a hiring manager at a software company in Kathmandu.
Here are the experience bullets from my resume: [paste your bullets here].
First, point out which bullets are vague or only describe duties. Then rewrite the 3 weakest ones to show a result or a number. Finally, tell me what is missing for a junior web developer role.
Show the output as a table with three columns: original line, improved line, reason.Two cautions for resumes:
- Do not paste your citizenship number, passport number, full home address, or phone number into any AI tool. The AI does not need them to improve your bullets.
- Do not let it invent numbers. If the AI writes “increased traffic by 40%” and that never happened, delete it. Only keep results you can explain in an interview.
If you do not have a resume yet, tell the AI your real details and ask it to draft one, then edit it yourself. Ask for a simple one-page format that works with applicant tracking systems, because many fancy templates with columns and icons get scrambled when a company uploads them.
Which framework should you use?
Start with TAG if you are unsure. If the answer comes back in the wrong shape, switch to CARE. If you want an expert opinion, use RACE. If you are fixing something you wrote, use RISE.
In practice, the frameworks overlap, and that is fine. Real prompts often borrow from two of them. The point is to cover who, what, why, and how, not to follow a formula perfectly.
Habits that fix most bad prompts
- Name your level. “Class 10 SEE student” and “BSc CSIT third semester” get very different answers.
- Name the exam or the audience. “NEB,” “SEE,” “Loksewa,” and “TSC” each have a different style.
- Paste your syllabus or chapter text when you can. The AI follows what you give it better than what it remembers.
- Say how long and in what format. “Under 200 words,” “table,” “bullet points,” “numbered steps.”
- Follow up. Your first prompt is a start. Say “make it shorter,” “add a numerical,” or “explain that part again like I am 12.” Most good results come from the second or third message.
What AI still gets wrong
AI tools are useful, but they make confident mistakes. Be careful with these:
- Nepal-specific facts. Laws, current affairs, recent policy changes, vacancy details and dates are where AI is most often wrong or out of date. Check them against an official source such as the Public Service Commission website or a recent textbook.
- Marking schemes. The AI does not know how your examiner will give marks. Compare its answers with your textbook and past papers.
- Numericals. Re-solve at least one step yourself. A wrong formula looks just as neat as a right one.
- Your homework. If you paste an answer in without understanding it, you lose the practice that the assignment was for. Use AI to explain, quiz you, and give feedback. Do the thinking yourself.
A fill-in template to save
Keep this in your phone notes and change the brackets each time:
I am a [your level] student in Nepal preparing for [exam].
[What you want the AI to do] on [topic or chapter].
Give me the answer as [format] in [length].
Use [language] and include [examples, MCQs, or a table].A good prompt is usually only four or five lines. Spend one extra minute saying who you are, what you need and what the answer should look like, and the results improve a lot.
Pick one framework and try it today on a real chapter or a real resume line.
Published October 2026. Last verified: October 2026.