Welcome to SpyLogic
SpyLogic is an innovative application designed to explore defensive measures against prompt injection attacks targeting Large Language Models (LLMs). With generative AI becoming increasingly prevalent, understanding the weaknesses inherent to these models is crucial. SpyLogic aims to educate users in a fun and engaging manner about prompt injection attacks and effective defense strategies.
Key Features
-
Two Interactive Modes:
- Story Mode: Go undercover to execute prompt injection attacks against ScottBrewBot, a clever but flawed generative AI bot. Progress through levels by extracting secrets while learning about LLMs and defensive measures.
- Sandbox Mode: Activate various prompt injection defense measures to create a personalized security system against AI threats.
-
Built with OpenAI API: Utilize the capabilities of the OpenAI API to effectively simulate and defend against prompt injection attacks.
-
Easy Setup: Quick installation guide and Docker support for local hosting or deployment on personal infrastructure.
Benefits
- Learn about the intricacies of prompt injection and LLM vulnerabilities in an interactive environment.
- Strengthen your understanding of AI security measures and prompt engineering through practical application.
- Participate in an open-source project that encourages contributions and enhancements from the developer community.
Highlights
- Written in TypeScript, ensuring strong typing and better tooling during development.
- Comprehensive documentation and community support for users and contributors.