Detects user input passed to LLMs without Unicode normalization, allowing RTL overrides, zero-width joiners, and homoglyph attacks. Use when writing code that passes user-supplied strings to an LLM, constructs prompts from external data, or processes text that may contain Unicode control characters, homoglyphs, or invisible formatting characters.
