A LinkedIn-based LLM jailbreak game where users attempt to bypass Gandalf's system prompt through comments on a LinkedIn post.

Inspired from the great Gandalf Lakera

This project creates an interactive game that:

• Monitors a specific LinkedIn post for comments
• Uses an LLM (playing as Gandalf) to respond to user attempts
• Allows each user only one attempt to bypass the system prompt
• Automatically responds to new comments
• Tracks user attempts to prevent multiple tries

1. Clone the repository:

git clone https://github.com/lucebert/gandalf-linkedin.git
cd gandalf-linkedin

2. Install dependencies using Poetry:

poetry install

3. Set up environment variables:

cp .env.example .env

Run the application:

poetry run python -m gandalf_linkedin.main

poetry run pytest

This project is released into the public domain under The Unlicense. See the LICENSE file for details.