Day 1 of 760 minCipher Cadet
Day 1: Orientation and your first security prompt
Taught by Mr Elite Updated
By the end of today you can
- Explain in plain words what a prompt and a model are
- Set up a notes file for recording your prompts and answers
- Write a prompt using role, task and format
- Judge whether an AI answer to a security question is clear enough to use
You have probably already typed something into an AI chat and got an answer that was vague, too long, or pitched at the wrong level. In security that costs you time. Ask badly about a term like firewall or phishing and you get a wall of text that assumes you already know the subject.
Today fixes that. You will learn what you are actually talking to, set up a notes file so your work is saved, and write a prompt that gets a short, clear security explanation. It is a small skill, and the next six days build on it, both for using AI well and for understanding how it gets tricked.
What this course covers
This course has two halves. The first three days teach you to use an AI assistant for security study and work without bad habits. The last four teach you how AI chatbots get attacked, so you can recognise the problems and reduce the risk.
Here is the route. Day 2 covers prompts for logs, code and reports. Day 3 covers keeping secrets out and checking answers. Days 4 to 6 cover prompt injection, system prompt leaks and jailbreaks. Day 7 is a capstone where you test a small assistant and write a defender report.
Each day follows the same shape and takes about an hour. I explain the idea, I show you a worked example, you try it yourself, and then you finish a short mission before the next lesson. You need three free things today: a web browser, a plain text editor, and access to any free AI chat assistant. You do not need to install anything special.
What a prompt and a model are
An AI chat assistant is a program built on a model. The model has learned patterns from huge amounts of text, and it works by predicting a sensible continuation of whatever text it is given. It does not look things up like a search engine, and it does not understand your situation unless you tell it.
What you type is the prompt. The model reads your prompt and writes a reply. Better input gives a better reply, because the prompt is nearly all the model knows about what you want.
This explains something that confuses many beginners. Ask the same short question twice and you may get two different answers, one for a student and one for an engineer. The model is guessing who you are and what you need. Every detail you leave out is a guess it makes for you, and guesses are where vague or wrong-level answers come from.
- Prompt
- The text you give an AI so it can produce a reply.
- Model
- The trained system that reads the prompt and generates the reply.
- Assistant
- The chat product wrapped around a model, such as the free tool you will use.
- Hallucination
- A confident answer that is wrong or invented. We deal with this properly on Day 3.
Quick check
Which statement best describes how a model produces a reply?
A model generates text from patterns, so it only knows what your prompt and its training give it.
Set up your notes file
You will keep a record of prompts and answers throughout the course. Professionals do the same, because a prompt that worked once is worth reusing. Let me get you set up before we touch the AI.
Day 1 notes - prompts and answersRole, task and format
A weak prompt is one short question. A stronger prompt gives the model three things. These are the same three I want you to reach for every time.
- Role
- Who the AI should act as, for example a patient security trainer.
- Task
- The specific thing you want done, including who it is for.
- Format
- The shape of the answer: length, layout, reading level.
Compare two prompts for the same topic. The first is what most people type. The second adds all three parts.
what is phishingYou are a patient security trainer. Explain what phishing is to a complete beginner. Use plain English, no jargon, in four short lines, then give one real-world example.Look at how each part earns its place. The role sets the tone, so you get patience instead of a dry textbook entry. The task names the topic and the reader. The format stops the answer running to ten paragraphs. If you remove any one part, the model fills the gap with its own guess.
Quick check
In the prompt 'You are a patient trainer. Explain DNS in three lines for a beginner', which part sets the format?
'In three lines' controls the shape and length of the answer, which is the format.
Watch me do it: your first prompt
Now I will run a prompt end to end. I will use the term 'two-factor authentication', and you will do the same with it so you can compare, then use your own term in the mission. Open any free AI chat assistant in your browser.
You are a patient security trainer. Explain what two-factor authentication is to a complete beginner. Use plain English, no jargon, in four short lines, then give one everyday example.Your wording will differ from mine, and that is normal. A good reply has roughly this shape, though. It is a guide to the shape, not text to copy.
Two-factor authentication means proving who you are in two different ways.
The first is something you know, like a password.
The second is something you have, like a code sent to your phone.
Even if a thief steals your password, they still cannot get in.
Example: a bank asks for your password, then a code from its app.What just happened: you gave the model a role, a clear audience and a shape, and it produced something you can actually read. Check the answer yourself. Does it say you prove who you are in two different ways? If it does, the answer is on track. If it says something odd, note that too, because Day 3 is about checking.
Your own practice
There is no lab today, so the practice happens on your own machine in your notes file. Pick one term from this list: malware, password manager, VPN, encryption. Write a prompt with role, task and format, run it, and record the result.
[ROLE] You are a patient security trainer.
[TASK] Explain what a password manager is to a complete beginner.
[FORMAT] Plain English, no jargon, in four short lines, then one example.If it did not work
Typing a vague one-line question and accepting whatever comes back
Add a role, the audience and a format. Even one extra sentence usually improves the answer.
Starting from an old chat full of earlier instructions
Open a new chat for each experiment so earlier text does not change the result.
Trusting the first answer without reading it
Read it for anything that sounds wrong and note it. Models can sound confident and still be mistaken.
Saving the file as day1.txt.rtf or similar
Use plain text mode and turn on file extensions so you can confirm the name.
Recap
- A prompt is the text you give; a model predicts a reply from it
- The model only knows what your prompt and its training give it
- Role, task and format make a prompt much clearer
- Naming the audience controls the reading level
- Iterate with short follow-ups instead of starting again
- Keep a notes file of prompts and answers, and never paste private data
Cheat sheet
Prompt- The input text you give the AI
Model- The trained system that generates replies
Role- Who the AI should act as
Task- What you want done and for whom
Format- Length, layout and reading level of the answer
Iterate- Refine the answer with a short follow-up message
Before the next day
Mission: Explain a firewall in five lines
Use the role, task and format pattern to get a clear beginner explanation of a firewall, then record it. This proves you can control an AI answer and builds the notes habit you will use all week.
About 30 min
You are done when: A saved day1.txt containing your firewall prompt and the final five-line answer
Want more? Run the same prompt with a different role, such as a bank manager, and write two lines in your notes on how the answer changed.
Quiz
4 questions. Your score is saved with your progress.
1. Why does a vague prompt often give a poor security explanation?
The prompt is nearly all the model knows about what you want, so missing detail leads to guesswork.
2. Which prompt best uses role, task and format?
It names a role, a task with an audience, and a length and style.
3. The answer to your prompt is far too long. What is the best next move?
Iterating with a short follow-up is quick and keeps the useful context.
4. Why do we keep a notes file of prompts and answers?
A record lets you reuse what worked and spot patterns in what went wrong.
Finish the day
Mark the day complete to save your progress.
Next: On Day 2 you will apply role, context, task and format to real material: a log line and a short code snippet. Today's pattern becomes the base of every prompt you write from here.
Your notes
Private to this device. Write what you want to remember from today.