The CTFC Prompt Framework: From Vague Wishes to Precision Instruments
If you’ve ever typed “Help me write something good” and received bland, generic text, you’ve experienced the frustration of a bad prompt. The standard advice is “be more specific,” but that’s like telling a novice cook to “make it taste better.” You need a recipe. Here it is: the CTFC Framework.
⚡ CTFC at a Glance
C – Context
Who are you? What’s the situation? What does the AI need to know about your world?
T – Task
One clear action verb + object. “Summarise…”, “Brainstorm 10…”, “Draft an email…”
F – Format
The shape of the answer: bullet points? table? email? code? Specify it.
C – Constraints
Guardrails: “Do not invent statistics”, “Under 150 words”, “If unsure, say so.”
Pillar 1: Context – The “Where Am I?” Information
Context answers: Who are you? What are you trying to achieve? What do I need to know? The richer the context, the more tailored the output.
Without context: “Write an email to my team.”
With context: “I’m a marketing manager at a mid‑size company, announcing a new flexible working policy to my team of 12. The tone should be excited but professional. It’s a Monday morning.”
Suddenly, the AI can picture the scene. Give it: your role, audience, background, and tonal cues.
Pillar 2: Task – The Action Verb and Object
The task is the core instruction. Use a strong, clear verb. Keep it singular; if you need multiple actions, use a numbered list.
Weak: “I was wondering if you could help me think about some ideas regarding…”
Strong: “Brainstorm 10 ideas for a sustainable side hustle under £100.”
Pillar 3: Format – The Shape of the Answer
Without a format, the AI defaults to a paragraph. You can ask for bullet points, tables, code blocks, scripts, markdown, word counts – anything that gives the output a mould to fill.
Examples: “Create a markdown table with columns Pros and Cons.” “Write a 150‑word email in a warm, professional tone.”
Pillar 4: Constraints – Guardrails and Hallucination Killers
Constraints tell the AI what not to do and how to handle uncertainty. This is your strongest defence against made‑up facts.
Powerful constraints: “Do not invent statistics.” “If you don’t know, say so.” “Only use information from the attached document.” “After answering, rate your confidence from 1‑5.”
A Complete CTFC Prompt, Built Step by Step
Let’s turn a vague request into a precision instrument.
Vague request: “Help me with a cover letter.”
CTFC version:
- C – Context: I’m applying for a junior software developer role at a fintech startup. I have a CS degree and a 6‑month internship at a bank. The job ad emphasises Python, cloud, and a ‘can‑do attitude’.
- T – Task: Write a 3‑paragraph cover letter.
- F – Format: Paragraph 1: introduction and interest. Paragraph 2: how my experience matches requirements. Paragraph 3: closing with a call to action. Confident but not arrogant tone.
- C – Constraints: Do not invent skills I didn’t mention. If you refer to the company, use ‘[Company Name]’. Keep under 200 words. After the letter, list any assumptions you made about my experience.
This prompt gives the AI everything it needs to produce a tailored, honest draft – and it’ll even flag its guesses. That’s the difference between hoping for the best and engineering the result.
Try It Now: Your First CTFC Prompt
Fill in the blanks and paste into any AI chatbot:
Task: [Verb + object]
Format: [e.g., bullet points, email, table]
Constraints: Do not [list]. If unsure, [say so/ask].
Example: “Context: I’m a freelance designer proposing a logo to a coffee shop owner who wants a vintage hand‑drawn feel. Task: Draft a 150‑word project proposal email. Format: Professional but warm, with subject line, approach, timeline, call to action. Constraints: No pricing. If you assume preferences, flag them.”
Now you have the framework. Ready to use it on ready‑made prompts?
🔎 Level Up: Once CTFC feels natural, learn the Contrarian Edit – a technique that forces the AI to check its own work and catch errors you might miss.