https://github.com/cx330blake/spell-whisperer
https://github.com/cx330blake/spell-whisperer
hacking llm pentesting prompt-injection redteam
Last synced: over 1 year ago
JSON representation
- Host: GitHub
- URL: https://github.com/cx330blake/spell-whisperer
- Owner: CX330Blake
- License: mit
- Created: 2025-03-09T12:14:52.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-03-19T07:04:08.000Z (over 1 year ago)
- Last Synced: 2025-03-19T07:23:15.213Z (over 1 year ago)
- Topics: hacking, llm, pentesting, prompt-injection, redteam
- Language: TypeScript
- Homepage: https://spell-whisperer.cx330.tw
- Size: 4.17 MB
- Stars: 24
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Spell Whisperer
- [Disclaimer](#disclaimer)
- [Introduction](#introduction)
- [Deploy](#deploy)
- [Deploy on Vercel](#deploy-on-vercel)
- [Deploy on your own](#deploy-on-your-own)
- [Customize the challenges](#customize-the-challenges)
## Disclaimer
This repository shows the basic ideas of hacking LLMs, but it's for educational purposes only. I am not responsible for any misuse of this repository.
## Introduction
Spell whisperer is a prompt injection challenge based on Grok API (of course, you can change it to any other API like OpenAI API). There's five challenges now, probably more in the future.
## Deploy
### Deploy on Vercel
By clicking the following button, you will clone a repo from here and deploy your own app on Vercel.
[](https://vercel.com/new/clone?repository-url=https://github.com/CX330Blake/Spell-Whisperer)
### Deploy on your own
1. Clone/Fork this repository
2. Install all the dependencies
```bash
npm install
```
3. Export your Grok/OpenAI API Key using this command
```bash
export XAI_API_KEY=YOUR_API_KEY
```
4. Run `npm build && npm start` to start the server
## Customize the challenges
You can simply customize them by modifying the `/src/app/api/challenge/challenges.json`. You'll see everything you need there.
## Star History
## TODO
1. Leaderboard
2. Challenges page refactor & add scoring system
3. Add flag prefix
4. Markdown response support
5. Other challenges (not just flag stealer)
6. Show others solutions