Fixed examples. They teach a structure you can reuse with your own facts.
Studying: Exam prep
Rough: help me study for my calculus exam
Goal: Help me study for a calculus exam.
Context: I am comfortable with algebra and new to limits and derivatives.
Constraints: Teach one idea at a time. Wait for my answer before showing a solution.
Desired output: A short explanation, one practice problem, then feedback on what I tried.
Name the topic, your current level, and how you want to practice.
Explaining: Difficult concept
Rough: explain photosynthesis
Goal: Explain photosynthesis so I can teach it to a younger sibling.
Context: I remember that plants use sunlight, but I mix up the inputs and outputs.
Constraints: Use everyday language first. Skip the history of the discovery.
Desired output: A short explanation, one analogy, and three questions I can use to check understanding.
Say who the explanation is for and what part is already confusing.
Writing: First draft
Rough: write a blog post about remote work
Goal: Draft a blog post about remote work for first-time managers.
Context: The audience is skeptical and has a team of eight people.
Constraints: About 700 words. No clichés about pajamas. Use one concrete weekly routine.
Desired output: A headline, the draft, and a one-sentence note on the audience it fits.
A topic is not a brief. Add the reader, the length, and one thing to avoid.
Rewriting: Clearer prose
Rough: make this better: We should circle back and leverage synergies.
Goal: Rewrite this sentence in plain language: "We should circle back and leverage synergies."
Context: It is for an email to a client who was not in the meeting.
Constraints: Keep the meaning. Do not add new promises.
Desired output: One replacement sentence and a one-line note on what was vague.
Paste the exact text and say what must stay the same.
Summarizing: Meeting notes
Rough: summarize this
Goal: Summarize the notes I paste next.
Context: I need to send the summary to someone who missed the meeting.
Constraints: Keep decisions and owners. Drop small talk.
Desired output: Five bullets, then a short list of open questions.
Say what the summary is for, or the model has to guess the audience.
Brainstorming: Options
Rough: give me startup ideas
Goal: Suggest small software ideas I could test in two weeks.
Context: I can code and I talk to independent bookkeepers every week.
Constraints: No marketplace ideas that need both sides on day one. Give 8 ideas.
Desired output: For each idea, the user, the painful job, and the smallest test.
Constraints make a brainstorm usable. "Give me ideas" has nowhere to land.
Coding: Implementation
Rough: write a python script to rename files
Goal: Write a Python script that renames files in one folder.
Context: The files look like IMG_1234.JPG and I want 2024-trip-01.jpg style names.
Constraints: Python 3. Do not delete files. Show a dry-run mode first.
Desired output: The script, how to run it, and what the dry run prints.
Name the language, the input shape, and what the code must not do.
Debugging: A failing test
Rough: my code is broken fix it
Goal: Find why this function returns the wrong total.
Context: I will paste the function, the input, the actual result, and the result I expected.
Constraints: Explain the cause before rewriting. Change as little as possible.
Desired output: The likely cause, a small patch, and one test I can run.
Include the symptom and the expected result. "Fix it" hides both.
Planning: A week of work
Rough: plan my week
Goal: Plan a work week for a product launch on Friday.
Context: I have about four focused hours a day. The launch email and the bug list are both unfinished.
Constraints: Put the launch-critical work first. Leave Wednesday afternoon empty.
Desired output: A day-by-day plan with one priority and one thing I can drop if time slips.
A plan needs a deadline, a limit, and what matters most.
Research: A comparison brief
Rough: research standing desks
Goal: Help me compare standing desks for a small apartment.
Context: I work from a laptop and I have about $400.
Constraints: Do not assume I want a brand. Separate facts you are sure of from things I should verify.
Desired output: A short comparison of desk types, what to measure in my room, and questions to ask a seller.
Ask for the decision you are making, and for a clear line between known and uncertain.
Job help: Resume bullet
Rough: fix my resume
Goal: Rewrite one resume bullet for a customer support role.
Context: The bullet is "Helped customers with problems." I also trained two new hires.
Constraints: Keep it truthful. Do not invent metrics.
Desired output: Two alternative bullets and a note on which one is stronger if I have no numbers.
Ask for one section at a time, and forbid invented achievements.
Email: A difficult note
Rough: write an email saying the project is late
Goal: Draft an email telling a client the project will be one week late.
Context: The delay is a vendor slip, not a scope change. We can still deliver the original features.
Constraints: Be direct and calm. Do not blame a named person. Offer one next step.
Desired output: A subject line and a short email I can edit.
Give the relationship, the fact, and the tone. Then edit the draft yourself.
Business ideas: A first offer
Rough: how do I make money freelancing
Goal: Help me shape a first freelance offer.
Context: I am good at spreadsheet cleanup for small shops and I can start next month.
Constraints: One offer, not a list of careers. Assume I have no audience yet.
Desired output: Who it is for, what is included, what is not, and how I could find the first three conversations.
Narrow the skill and the starting point before asking for a business plan.
Comparing options: A choice
Rough: which is better chatgpt or claude
Goal: Help me choose where to draft a client proposal this week.
Context: I already use both ChatGPT and Claude. The proposal is two pages and the client wants plain language.
Constraints: Do not declare a universal winner. Compare them only for this job.
Desired output: What to try in each chat, and the question I should use to judge the drafts.
Compare tools for a specific job. A blanket "which is better" question stays vague.