Chapter 0
This is the mdbook for the prompt injection lab.
For hints checkout the hints page. (or ask the HackSoc Board)
For the challenges check out the challenges chapter.
Guide
Imagine what the system prompt could be.
What is it trying to stop outputting?
How might you try and get the system prompt?
Are there any ways to communicate information without explicitly typing the words as you can read them?
Are there any ways to control the output of the LLM more explicitly?
Challenges
Start Here:
https://prompting.ai.immersivelabs.com/
If you finish that try here:
https://app.grayswan.ai/arena/challenge/onboarding-quest-1
Want to learn more:
https://cheatsheetseries.owasp.org/cheatsheets/LLM_Prompt_Injection_Prevention_Cheat_Sheet.html
https://portswigger.net/web-security/llm-attacks