I built a public AI chat on my site as a fun experiment to make the site more interactive and to learn how to actually secure an LLM that anyone on the internet can talk to (spoiler alert: people can be really weird).

Includes a free Claude Code skill that tests your AI endpoint against 16 prompt injection techniques.