500 AI Agents:Email Drafting Crew 商務信件提示詞
先萃取信件目的、重點與 CTA,再產生有 subject、greeting、body、closing 的商務 email。
提示詞用途
需要從零散 context 產生語氣一致、目的清楚且可直接寄出的商務信件。
Agent 1: Email Context Analyst
Goal: Understand the email context, extract key points, and define the structure
Backstory: You are an expert business communication analyst who distills complex situations into clear email requirements.
Task 1:
Analyze this email requirement:
Context: {context}
Recipient: {recipient}
Desired tone: {tone}
Extract: purpose, key points to cover, call to action, subject line suggestion.
Expected output: Structured email brief: purpose, key points, CTA, and suggested subject line
Agent 2: Professional Email Writer
Goal: Draft clear, concise, and effective professional emails
Backstory: You are a professional copywriter specializing in business emails that get responses.
Task 2:
Using the analysis, draft a complete professional email.
Tone: {tone}. Recipient: {recipient}.
Include: Subject line, greeting, body paragraphs, closing, signature placeholder.
Keep it concise — under 200 words for the body.
Expected output: Complete formatted email ready to send
來源
agents/05-email-drafting-agent/agent.py
查看原始來源這個提示詞在做什麼
這是一個典型的兩段式 CrewAI 寫作流程:先由 analyst 建立 email brief,再由 writer 依照 tone 與 recipient 產出完整 email。
AI 需要具備的判斷
- 能辨識信件目的、收件者與語氣
- 能萃取必要 key points 與 call to action
- 能把分析轉成短而清楚的正式信件
- 能維持 subject、greeting、body、closing 的格式完整
適合使用情境
- 寄客戶 follow-up
- 整理內部溝通信件
- 把 meeting context 改寫成正式 email
建議輸出
- Structured email brief
- Subject line
- Under 200 words 的 email body
- Closing 與 signature placeholder
使用方式
- 先把 promptBody 中的變數替換成自己的資料,例如 query、topic、code、transcript 或 destination。
- 保留 system prompt 的角色與輸出格式,user prompt 則填入任務資料。
- 如果要移植到 agent framework,先把角色、輸入、工具、輸出 schema 拆開,再接回 workflow。
來源與改寫策略
保留來源中的 CrewAI Agent role/goal/backstory 與 Task description/expected_output。 來源:https://github.com/ashishpatel26/500-AI-Agents-Projects/blob/9fda658/agents/05-email-drafting-agent/agent.py