Prompt injection is when a user tricks the model into ignoring prior instructions revealing system prompts, disabling safeguards or acting outside intended boundaries.
I first saw it live during DEF CON (31) finals and have since seen it exploited in bug bounty reports and research.
This is a small proof-of-concept that works like an “AI firewall”
detecting injection attempts before they reach your LLM with almost no added latency.
Blog post: https://blog.himanshuanand.com/posts/2025-08-10-detecting-llm-prompt-injection/
Demo/API: https://promptinjection.himanshuanand.com/
fast, API friendly and has a UI for testing bypass attempts (For CTF enthusiastic people like me). Feedback and break attempts welcome.