Cipher CadetBeginner9 daysFree
Prompt Injection From Zero: Attack and Defend LLM Apps
In nine short days you will go from never having touched security to breaking, explaining and defending a simple AI app.
Taught by Mr Elite 9 of 9 days live Complete
I built this course for people who use AI tools and want to understand how they can be turned against their owners. Prompt injection is the most common weakness in apps built on large language models (LLMs), the engines behind chatbots. You do not need any security background. If you can use a browser and type into a text box, you can start.
The course runs for nine days, one lesson a day, about 45 to 75 minutes each. Every day has one objective and one mission that produces something I can check: a screenshot, a note or a short write-up. You practise in SecurityElites browser labs, so everything you attack is a practice target built for the purpose. Nothing is aimed at a real system. The last day is a capstone where you use everything and learn what to study next.
By the end you can
- Explain in plain words what a prompt, a system prompt and prompt injection are.
- Perform a basic direct prompt injection against a practice chatbot and describe why it worked.
- Explain the difference between direct and indirect injection, and carry out an indirect one in a lab.
- Recognise how injection becomes more serious when an AI has tools, memory or a knowledge base.
- Describe the main extraction and bypass tricks, including prompt leaking and encoding.
- Apply layered defences and write a short, clear finding report on a prompt injection issue.
Who this is for
- Complete beginners in security who want a clear first step into AI security.
- Developers, analysts and curious users who build or use chatbots and want to know how they fail.
- Anyone who wants a practical, hands-on grounding before moving to harder AI hacking labs.
Before you start
- Nothing beyond being able to use a computer and a web browser.
- A text editor or notes app for keeping your mission notes.
- About an hour a day for nine days.
- A modern web browser (Firefox or Chrome)
- A free SecurityElites account
- A plain text editor or notes app
- A screenshot tool (built into your operating system)
Syllabus
One day at a time. Finished days get a tick. Days that are not out yet show when they unlock.
Module 1 Foundations: How LLM Apps Work and Break
Understand how LLM apps are built and perform your first direct prompt injection.
- 1Orientation and Your First Chat With a Bot Under TestBy the end of today you can describe what this course teaches, set up your workspace and make a practice chatbot say something its owner did not intend.Lab: AI Prompt Injection
- 2System Prompts and Why Instructions Are Not WallsBy the end of today you can explain the difference between a system prompt and user input, and why the model cannot truly tell them apart.Practice | Lab: AI System Prompt Extraction
- 3Direct Injection: Conflicting InstructionsBy the end of today you can craft a direct prompt injection that exploits a conflict between the app's rules and your request, and explain why it worked.Practice | Lab: AI Instruction Conflict
Module 2 Indirect Injection: Attacks Hidden in Content
Understand how hostile instructions arrive through documents, images and data the AI reads.
- 4Indirect Injection: When the Document Is the AttackerBy the end of today you can explain indirect prompt injection and carry out one against a document summariser.Lab: AI Indirect Prompt Injection
- 5Poisoned Knowledge: RAG and Retrieved ContentBy the end of today you can describe how retrieval-augmented generation (RAG) feeds outside text to a model and show how a planted entry changes answers.Practice | Lab: AI RAG Poisoning
- 6Hidden Text: Images and Invisible CharactersBy the end of today you can show how instructions hidden in images or invisible text reach a model unnoticed, and apply what Days 4 to 5 taught.Review | Lab: AI Visual Prompt Injection
Module 3 Impact, Defence and Capstone
Connect injection to real harm, apply layered defences, and complete a full attack and report.
- 7When the AI Can Act: Tools, Memory and Real HarmBy the end of today you can explain why injection becomes far more dangerous when an AI has tools or memory, and demonstrate a poisoned issue steering a coding agent.Practice | Lab: Poisoned GitHub Issue | Lab: AI Memory Poisoning
- 8Defending LLM Apps in LayersBy the end of today you can list and apply layered defences against prompt injection and judge which ones a given weakness needs.Practice | Lab: Coding Agent Blast Radius
- 9Capstone: Attack, Report, Defend, and What NextBy the end of today you can run a full prompt injection test on a practice target, write it up clearly and recommend fixes, then choose your next learning path.Capstone | Lab: AI Agent Hijacking
The final project
You will attack a practice AI app of your choice from the lab list, using at least two techniques from the course, and write a one-page report covering what you did, why it worked and how you would fix it. You then review your own report against a defence checklist and pick your next learning steps.
Questions people ask
Do I need to know how to code?
No. Every lab runs in your browser and you attack by typing text. I explain each term the first time it appears, so you can follow without any programming.
Is it legal to try these attacks?
Yes, because you only attack the SecurityElites practice labs, which exist for this purpose. Never try these techniques on a real chatbot or service you do not own or have written permission to test.
How long does each day take?
Plan for 45 to 75 minutes, including the lab and the mission. Some labs may take you longer on your first try, and that is normal.
Will this make me a professional AI security tester?
It gives you a solid foundation, not a finished career. By Day 9 you will understand the core ideas and have practical experience, and I point you to the next topics so you can keep going.