什么是 邮件分选助手？

邮件分选助手是 Openclaw 技能库中的一个专业工具，旨在帮助 AI 智能体处理复杂的通信流。通过应用系统化的分选框架，该技能使智能体能够按紧急程度对邮件进行排序，提取隐藏的截止日期，并检测邮件链中微妙的情绪转变。它将杂乱的收件箱转变为结构化的待办任务列表，确保高优先级沟通得到立即关注，同时妥善归档常规通知。利用这些 Openclaw 技能，可以实现更高效、更专业的数字化沟通。

下载入口:https://github.com/openclaw/skills/tree/main/skills/1kalin/afrexai-email-triager

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

邮件分选助手 应用场景

• 处理每日海量邮件，将紧急请求与信息更新分开。
• 针对客户咨询和内部请求，自动撰写专业且语气匹配的回复草案。
• 识别经理或高管被加入邮件链的关键升级情况。
• 从冗长的邮件链中提取具体的任务和截止日期，并以清单格式呈现。
• 识别同一发送者多封邮件中的沟通模式，建议进行如预约电话等更深层次的干预。

1. AI 智能体分析来信的内容、发送者和元数据。
2. 每封邮件被分配到特定的分选类别，如“紧急行动”、“需要行动”、“供参考”、“授权”或“归档”。
3. 应用智能信号检测升级情况、情绪变化或隐藏的会议请求。
4. 为每封邮件生成结构化输出，包括简要总结和明确的下一步操作。
5. 对于需要行动的项目，该技能根据特定的简洁和语气模仿规则，生成直接且符合语境的回复草案。
6. 批量处理多封邮件，以识别跨邮件链的模式和系统优先级。

邮件分选助手 配置指南

要部署邮件分选助手，请将 SKILL.md 定义集成到 AI 智能体的系统提示词或工具配置中。确保智能体拥有阅读邮件标题和正文内容的权限。

# Openclaw 技能集成的示例配置
openclaw skills:add email-triager

您还可以将其与针对法律或医疗等专业行业的领域特定上下文包配对，以细化分选逻辑。

邮件分选助手 数据架构与分类体系

邮件分选助手将其输出组织成标准化模式，兼顾人类可读性和程序化处理：

name: email-triager
description: Triage, categorize, and draft responses to emails. Sorts by urgency, flags action items, and generates context-aware reply drafts.

Email Triager

When asked to triage, sort, or process emails, follow this system.

Triage Categories

Assign every email exactly one category:

Triage Output Format

For each email, produce:

[ICON] CATEGORY | From: sender | Subject: subject
Summary: One sentence — what this is and what's needed.
Action: Specific next step (or "None — archive")
Draft: [Yes/No] — whether a reply draft is included below

Draft Response Rules

Generate a reply draft when:

• Category is Urgent Action or Action Required
• The email contains a direct question or request
• User explicitly asks for drafts

Draft style:

• Be direct. Open with the answer or decision, not "Thank you for your email."
• Mirror their tone. Formal email gets formal reply. Casual gets casual.
• Keep it short. Most replies should be 2-5 sentences.
• End with clarity. What happens next? Who does what by when?
• Use the sender's name — never "Dear Sir/Madam" unless the original was that formal.

Batch Processing

When given multiple emails:

1. Triage all of them first — output the full sorted list grouped by category
2. Then provide drafts for Urgent Action and Action Required items
3. Highlight any patterns ("3 emails from the same client — might want a call instead")

Smart Signals

Flag these automatically:

• Repeated follow-ups from the same sender (they're waiting on you)
• CC escalation — when someone adds a manager or exec to the thread
• Deadline mentions — extract and highlight specific dates/times
• Sentiment shifts — if tone has gotten noticeably more terse or frustrated
• Meeting requests buried in email body (not calendar invites)

Action Item Extraction

Pull out discrete action items from emails:

• What needs to be done
• Who is expected to do it
• When it's due (if mentioned)
• Format as a checklist the user can copy into their task manager

When Not Triaging

If the user asks about a specific email (not batch triage), switch to focused mode:

• Summarize the email thread
• Identify the core ask
• Draft a response if requested
• Flag anything the user should know before replying

Pair with an industry context pack for domain-specific email handling (legal, healthcare, finance, etc.) at https://afrexai-cto.github.io/context-packs