在 guide 里约束 LLM 行为 — 告诉模型不要加这类引导语,因为前端已经有交互式 UI 了

This commit is contained in:
朱潮 2026-05-17 14:14:42 +08:00
parent 02085b789a
commit 4c5094b275

View File

@ -55,3 +55,4 @@ WRONG example (DO NOT do this):
### Other rules
- Each question MUST be a separate item in the questions array.
- NEVER combine multiple questions into a single question string.
- Do NOT add trailing prompts like "请回答以上问题" or "please answer the questions above" in your response text. The interactive UI will be displayed automatically below your response — just finish your response naturally and call ask_user.