The R-C-T-F recipe | Master AI Automation in 4 hours Master AI Automation in 4 hours Course About Ayush Modules Sample chapter Toolbox The Microcap Minute Classroom / Module 02: Talking to Machines: Prompting / Chapter 2 The R-C-T-F recipe Watch first, then read. Same lesson, your pace. What you will learn – The four ingredients present in nearly every great prompt – How role-setting changes vocabulary and depth – Why format instructions save the most time of all Here is a recipe simple enough to memorise today and use for the rest of your life: R-C-T-F . R, Role: who is answering? Start with: “You are a ___.” Telling the model who to be changes which patterns it reaches for. “You are a patient physics teacher for a 12-year-old” pulls up simpler words, analogies, encouragement. “You are a strict journal reviewer” pulls up scepticism and structure. Same brain, different hat. Roles also set standards : “You are a careful editor who hates filler words” gives the model something to be faithful to. C, Context: what does it need to know? The facts of your situation. Paste the actual material, the essay, the error message, the data, the constraints (“I have ₹500”, “deadline tomorrow”, “reader is my grandmother”). From Module 1: anything not on the desk gets guessed. Context is you putting things on the desk. Rule of thumb: if a human colleague would need it to do the job well, paste it. Don’t paste what they wouldn’t, irrelevant context dilutes attention. T, Task: what exactly should it do? One verb, made specific. Not “help with history” but “list the three main causes of World War I, each with a two-line explanation”. If your task has steps, number the steps. Models follow numbered lists remarkably well. Add constraints here too: word count, reading level, language, things to avoid (“no jargon”, “don’t invent statistics”). F, Format: what shape should the output take? The highest-leverage, most-skipped ingredient. Say exactly what you want back: “A markdown table with columns: Date | Event | Why it mattered” “JSON with keys title, summary, tags[]” “Five bullet points, max 12 words each” Format turns output from something to reformat into something to use directly , ready to paste into notes, spreadsheets, slides or code. Assembled You are an experienced maths tutor for Class 8 students. Context: my daughter understands fractions but freezes on decimals; below are her last three test questions (pasted). Task: create five practice problems that go from easy to hard, each teaching one specific decimal skill she got wrong. No calculators allowed. Format: numbered problems with answers upside-down-style at the very end, plus one-line hint per problem. Four lines. Zero guessing. This pattern works identically in Claude, Kimi, Qwen, ChatGPT, and later in this course, inside APIs and agents, where R-C-T-F becomes literally the system prompt. Try it yourself Pick something real you need this week (notes, email, study plan). Write it as R-C-T-F with each ingredient labelled. Run it. Then delete just the F section and run again, see how much cleanup time format saved you. Save the winning prompt to prompts/rctf-template.txt with the labels as placeholders. Key takeaways – Role sets standards and vocabulary; Context puts facts on the desk. – Task = one specific verb plus constraints; numbered steps work. – Format instructions convert output from draft to usable artifact. – R-C-T-F scales everywhere: chats today, system prompts in Modules 4-10. Download the exercise sheet (PDF) Module workbook (PDF) ← Prev: Why prompts fail Next: Show, don’t tell → Classroom / Module 02: Talking to Machines: Prompting / Chapter 2 The R-C-T-F recipe What you will learn – The four ingredients present in nearly every great prompt – How role-setting changes vocabulary and depth – Why format instructions save the most time of all Here is a recipe simple enough to memorise today and use for the rest of your life: R-C-T-F . R, Role: who is answering? Start with: “You
The R-C-T-F recipe
Written by
in