token-smuggling

From README

Token Smuggling / Unicode Injection (LLM01:2025) What this checks Detects user input passed to LLMs without Unicode normalization. Attackers embed RTL override characters, zero-width joiners, or homoglyphs to manipulate prompt structure, bypass keyword filters, or make malicious instructions appear legitimate. Vulnerable patterns User input interpolated into a prompt with no normalization step at the trust boundary Blocklists or keyword filters that compare against pre-normalized text, letting homoglyph variants pass Retrieved RAG content or tool outputs concatenated into a prompt on a path the sanitizer does not cover Comparisons on raw bytes where bidirectional overrides or zero-width characters can split or hide tokens Fix immediately Flag the vulnerable code, explain the risk, and suggest a fix establishing these properties.