Living note

Prompt Evaluation Before Production

A deliberately English-only mock note for checking the Chinese fallback experience.

当前笔记暂缺中文版本,正在显示英文内容。

A prompt change is a behavior change. Treat it like code: define expectations, run cases, and inspect regressions.

Minimum evaluation set

  • Common successful requests
  • Ambiguous or incomplete requests
  • Adversarial and policy-sensitive requests
  • Previously observed regressions