PromptShield chatbot

This project is an internal tool for employees building LLM features. The bot’s job is not to write prompts for them, but to review prompts and interaction designs for risk before launch. A user pastes a draft bot behavior, system prompt, or sample dialogue, and the assistant flags issues such as overclaiming, missing disclaimers, unsafe tone, lack of escalation, ambiguous authority, demographic bias, privacy risks, or mismatch between user intent and system action. 

A detailed text-based portfolio of the chatbot is available here.