模型提示词生成
A meta-prompt for generating stronger prompts from a rough task description, constraints, and expected output.
審校時間: 2026-03-18
Use this when you know the task you want the model to do but want the model to draft the actual prompt more systematically.
Prompt 模板
You are a prompt engineer.
Given the task below, write the best possible prompt for another language model.
Task:
{task}
Context:
- Audience: {audience}
- Output format: {output_format}
- Constraints: {constraints}
- Quality bar: {quality_bar}
Return:
1. The final prompt.
2. Why this prompt is structured this way.
3. Optional variants for stricter or faster output.正式使用前，請先替換模板變數並補充你的限制條件。
適用場景
- Turning rough task ideas into reusable prompts
- Building internal prompt libraries
- Improving prompt quality before automation
期望輸出
- Final prompt
- Reasoning for the structure
- Variants for different tradeoffs
可替換變數
What the downstream model should do
Example: Summarize support calls into CRM-ready notes
Who the output is for
Example: sales operations team
Expected output shape
Example: bullet summary plus JSON fields
Non-negotiable limits
Example: No invented details; preserve customer quotes exactly
What good output looks like
Example: concise, accurate, ready to paste into a CRM
優化建議
- Give the meta-prompt enough context to optimize for the right output.
- State the quality bar explicitly so the generated prompt is not generic.
- Ask for variants if you need a faster and a stricter version.
示例用法
Generate a prompt that turns bug reports into clean engineering triage notes.
預期效果
The model gives you a reusable prompt plus a rationale and variants you can test.