Normalize review evidence. Part of the How to Evaluate AI Tools cluster — keep outputs tied to a named owner and a measurable job.
Last updated: 2026-08-02
Prompt
Turn these notes into an evidence log with claim, source, date, and confidence: {NOTES} Usage
No invented sources.
Tips
Mark hearsay clearly.
Quality bar
Reject outputs that skip risks, invent metrics, or omit the next human action.
Related learning
Pair with tutorials in this cluster and the Learning Center.
FAQ
Safety
Redact secrets and customer PII before running.
When to retire?
Replace the prompt when edit rates stay high for two weeks.