Module prompt 

Prompt hardening utilities.

Functions

truncate_for_prompt

Truncate a string for safe embedding into LLM prompts.

wrap_untrusted

Wrap untrusted text in explicit XML-like tags to make the security boundary obvious to the model.