Cipher CadetBeginner7 daysFree
Prompt Engineering for Security Work: Use AI and Defend It
In seven days you will prompt an AI well for security tasks, then learn how those same prompts get attacked and how to defend them.
Taught by Mr Elite 2 of 7 days live A new day every day
I built this course for people who are new to security and want to use AI sensibly. Prompt engineering simply means writing instructions for an AI model so it gives you useful, accurate answers. In the first half of the week we use that skill on real security tasks: explaining logs, reading code, and drafting notes, while keeping secrets out of the conversation and checking what the model tells you.
In the second half I show you how prompts get attacked, always from the defender side. You will see system prompt leaks, injection and jailbreaks in the SecurityElites practice labs, and learn what a sensible defence looks like. Each day takes 45 to 75 minutes, includes one mission that produces something I can check, and builds on the day before. Everything runs on your own machine or in the named lab, never against anyone else's systems.
By the end you can
- Write clear prompts for security tasks using role, context, task and format.
- Redact secrets and sensitive data before pasting anything into an AI tool.
- Check AI answers for errors, invented facts and false confidence.
- Explain prompt injection, system prompt leaks and jailbreaks in plain words.
- Recognise each of these attacks in a lab and describe what went wrong.
- Write a defensive checklist for a prompt-based application.
- Review a small AI assistant design and recommend fixes.
Who this is for
- You are new to security and have no background beyond using a computer and a browser.
- You want to use AI assistants for security study or work without picking up bad habits.
- You want to understand how AI chatbots get tricked so you can spot and reduce the risk.
- You do not need to code, though you will copy and run a few simple commands.
Before you start
- Nothing beyond a computer, a modern browser and a free SecurityElites account.
- Access to any general AI chat assistant with a free tier.
- A text editor such as Notepad, TextEdit or VS Code for keeping notes.
- A modern web browser
- A free SecurityElites account for the labs
- Any free AI chat assistant
- A plain text editor for notes
Syllabus
One day at a time. Finished days get a tick. Days that are not out yet show when they unlock.
Module 1 Prompting for Security Tasks
Write good prompts, protect your data and verify what the AI says.
- 1Orientation and your first security promptBy the end of today you can write a basic prompt that gets an AI to explain a security term clearly.
- 2Prompts for logs, code and reportsBy the end of today you can use role, context, task and format to get useful help reading a log line and a short code snippet.Practice
- 3Keep secrets out and check the answersBy the end of today you can redact secrets from text before pasting it into an AI and spot at least two kinds of AI mistake.Practice | Lab: Secrets in the Context Window | Lab: Tests That Test Nothing | Unlocks 17 Nov
Module 2 How Prompts Get Attacked
Recognise injection, prompt leaks and jailbreaks and know how to defend against them.
- 4Prompt injection: when text becomes ordersBy the end of today you can explain prompt injection and demonstrate it against the practice sales bot.Lab: AI Prompt Injection | Unlocks 18 Nov
- 5System prompt leaksBy the end of today you can describe how a system prompt leaks and list three signs that a prompt is exposed.Practice | Lab: AI System Prompt Extraction | Unlocks 19 Nov
- 6Jailbreaks and how to spot themBy the end of today you can name two jailbreak styles, recognise them in a conversation and explain how a defender responds.Practice | Lab: AI Jailbreak | Lab: AI Refusal Suppression | Unlocks 20 Nov
- 7Capstone: defender report and next stepsBy the end of today you can test a small AI assistant for injection and prompt leakage and write a defender report with fixes.Capstone | Lab: AI Indirect Prompt Injection | Unlocks 21 Nov
The final project
You will take a small AI assistant design, test it in a lab for injection and prompt leaks, and write a one-page defender report. The report covers what you found, why it worked and the specific fixes you would make.
Questions people ask
Do I need to know how to code?
No. You will copy and run a few simple commands, and I explain each one. Nothing here needs programming experience.
Is it safe to try these attacks?
Yes, because every attack runs only in the SecurityElites practice labs or on your own setup. I never ask you to test a real third-party system, and you should not.
Which AI assistant should I use?
Any general free chat assistant will do. The skills you learn apply to all of them, so pick one you already have access to.
Will this make me a prompt injection expert in a week?
No, but you will understand the main ideas and be able to spot the common problems. It gives you a solid base to build on, and Day 7 points you to what to learn next.