Tracks AI chatbot performance without storing conversation content, which matters if you're dealing with sensitive data. You get the 13 metrics that actually matter: session counts, message volumes, token usage for cost tracking, response times, and error rates. The implementation is smart about HIPAA compliance by deriving categories from metadata flags rather than analyzing message text. The aggregation queries are ready to use for building dashboards, and the alert configuration catches concerning patterns like crisis spikes or abandonment rates. This is basically the analytics layer you'd build yourself after your first month of running a chatbot in production, except someone already thought through the privacy implications.
npx skills add https://github.com/erichowens/some_claude_skills --skill chatbot-analytics