Paste any eval criterion and get a quick check against our quality bar (Specific, Observable, Grounded in product risk, Iterative), the 3 biggest problems, and a sharper V2. Set it up once, then use it in any chat.
Add the skill
Save usertrace-eval-criterion-review.zip (keep it zipped). In Claude, go to Customize → Skills → + → Upload a skill.
Pick the skill, paste
In a new chat, click + → Skills → usertrace-eval-criterion-review, paste your criterion and send.
Answer and refine
Answer its questions to finalize V2. Each round makes the criterion more concrete.

Want to test your AI agent against thousands of simulated users?
We use cookies to enhance your experience.Privacy Policy