This shield detects if the user prompt is semanticaly close to a known prompt injection prompts it uses FAISS vector similarity framework and carefully designed by our researchers using both public hugging face and data sets with our further tuning