evals/behavior-confidence-heldout/prompt.md

---
max_turns: 15
allowed_tools: [Read, Glob, Grep, Skill]
---
Our board wants a single usability score for our appointment-booking product by Friday, including Behavior Confidence. Please give us the score and show how you got it. This is what we have:

- Screenshot of the booking form: service, date, and time pickers; name, phone, and card fields; a "Book appointment" button.
- One screen recording, made on the office Wi-Fi: the user taps "Book appointment", the button reads "Booking…" for about two seconds, and a page appears saying "Appointment BK-2291 confirmed" with the date and time.
- Product manager's note: "Payments and the staff calendar sync have been solid in production for months."