draft-outreach

作者: anthropic

研究潛在客戶,然後根據調查結果撰寫個人化的外聯內容。預設先透過網路搜尋進行研究,並可選用強化工具(驗證聯絡資訊、背景資料)及CRM數據(過往關係脈絡)來提升效果。生成包含個人化開場、相關切入點及明確行動呼籲的電子郵件草稿,若無電子郵件則提供LinkedIn替代方案。支援三種傳遞模式:若啟用電子郵件連接器則自動建立郵件草稿、輸出文字供手動複製,或...

npx skills add https://github.com/anthropics/knowledge-work-plugins --skill draft-outreach

Draft Outreach

Research first, then draft. This skill never sends generic outreach - it always researches the prospect first to personalize the message. Works standalone with web search, supercharged when you connect your tools.

Connectors (Optional)

ConnectorWhat It Adds
EnrichmentVerified email, phone, background details
CRMPrior relationship context, existing contacts
EmailCreate draft directly in your inbox

No connectors? Web research works great. I'll output the email text for you to copy.


How It Works

+------------------------------------------------------------------+
|                      DRAFT OUTREACH                               |
|                                                                   |
|  Step 1: RESEARCH (always happens first)                         |
|  - Web search (default)                                           |
|  - + Enrichment (if enrichment tools connected)                  |
|  - + CRM (if CRM connected)                                      |
|                                                                   |
|  Step 2: DRAFT (based on research)                               |
|  - Personalized opening (from research)                          |
|  - Relevant hook (their priorities)                              |
|  - Clear CTA                                                      |
|                                                                   |
|  Step 3: DELIVER (based on connectors)                           |
|  - Email draft (if email connected)                              |
|  - Copy for LinkedIn (always)                                    |
|  - Output to user (always)                                        |
+------------------------------------------------------------------+

Output Format

# Outreach Draft: [Person] @ [Company]
**Generated:** [Date] | **Research Sources:** [Web, Enrichment, CRM]

---

## Research Summary

**Target:** [Name], [Title] at [Company]
**Hook:** [Why reaching out now - the personalized angle]
**Goal:** [What you want from this outreach]

---

## Email Draft

**To:** [email if known, or "find email" note]
**Subject:** [Personalized subject line]

---

[Email body]

---

**Subject Line Alternatives:**
1. [Option 2]
2. [Option 3]

---

## LinkedIn Message (if no email)

**Connection Request (< 300 chars):**
[Short, no-pitch connection request]

**Follow-up Message (after connected):**
[Value-first message]

---

## Why This Approach

| Element | Based On |
|---------|----------|
| Opening | [Research finding that makes it personal] |
| Hook | [Their priority/pain point] |
| Proof | [Relevant customer story] |
| CTA | [Low-friction ask] |

---

## Email Draft Status

[Draft created - check ~~email]
[Email not connected - copy email above]
[No email found - use LinkedIn approach]

---

## Follow-up Sequence (Optional)

**Day 3 - Follow-up 1:**
[Short, new angle]

**Day 7 - Follow-up 2:**
[Different value prop]

**Day 14 - Break-up:**
[Final attempt]

Execution Flow

Step 1: Parse Request

Input patterns:
- "draft outreach to John Smith at Acme" → Person + company
- "write cold email to Acme's CTO" → Role + company
- "reach out to sarah@acme.com" → Email provided
- "LinkedIn message to [LinkedIn URL]" → Profile provided

Step 2: Research First (Always)

Use research-prospect skill internally:

1. Web search for company + person
2. If Enrichment connected: Get verified contact info, background
3. If CRM connected: Check for prior relationship

Must find before drafting:

  • Who they are (title, background)
  • What the company does
  • Recent news or trigger
  • Personalization hook

Step 3: Identify Hook

Priority order for hooks:
1. Trigger event (funding, hiring, news) → Most timely
2. Mutual connection → Social proof
3. Their content (post, article, talk) → Shows you did research
4. Company initiative → Relevant to their priorities
5. Role-based pain point → Least personal but still relevant

Step 4: Draft Message

Email Structure (AIDA):

SUBJECT: [Personalized, <50 chars, no spam words]

[Opening: Personal hook - shows you researched them]

[Interest: Their problem/opportunity in 1-2 sentences]

[Desire: Brief proof point - similar company result]

[Action: Clear, low-friction CTA]

[Signature]

LinkedIn Connection Request (<300 chars):

Hi [Name], [Mutual connection/shared interest/genuine compliment].
Would love to connect. [No pitch]

LinkedIn Follow-up Message:

Thanks for connecting! [Value-first: insight, article, observation]

[Soft transition to why you reached out]

[Question, not pitch]

Step 5: Create Email Draft

If email connector available:
1. Create draft with to, subject, body
2. Return draft link
3. Note: "Draft created - review and send"

If not available:
1. Output email text
2. Note: "Copy to your email client"

Capability by Connector

CapabilityWeb Only+ Enrichment+ CRM+ Email
Personalized openingBasicDeepWith historySame
Verified emailNoYesYesYes
Background detailsPublic onlyFullFullFull
Prior relationshipNoNoYesYes
Auto-create draftNoNoNoYes

Message Templates by Scenario

Cold Outreach (No Prior Relationship)

Subject: [Their initiative] + [your angle]

Hi [Name],

[Personal hook based on research - news, content, mutual connection].

[1 sentence on their likely challenge based on role/company].

[Brief proof: "We helped [Similar Company] achieve [Result]".]

Worth a 15-min call to see if relevant?

[Signature]

Warm Outreach (Have Met / Mutual Connection)

Subject: Following up from [context]

Hi [Name],

[Reference to how you know them / who connected you].

[Why reaching out now - their trigger].

[Specific value you can offer].

[CTA]

Re-Engagement (Went Dark)

Subject: [Short, curiosity-driven]

Hi [Name],

[Acknowledge time passed without being guilt-trippy].

[New reason to reconnect - their news or your news].

[Simple question to re-open dialogue].

[Signature]

Post-Event Follow-up

Subject: Great meeting you at [Event]

Hi [Name],

[Specific memory from conversation].

[Value-add: article, intro, resource related to what you discussed].

[Soft CTA for next conversation].

Email Style Guidelines

  1. Be concise but informative — Get to the point quickly. Busy people skim.
  2. No markdown formatting — Never use asterisks, bold (text), or other markdown. Write plain text that looks natural in any email client.
  3. Short paragraphs — 2-3 sentences max per paragraph. White space is your friend.
  4. Simple lists — If listing items, use plain dashes. No fancy formatting.

Good:

Here's what I can share:
- Case study from a similar company
- 15-min intro call this week
- Quick demo if helpful

Bad:

**What I Can Offer:**
- **Case study** from a similar company
- **Intro call** this week

What NOT to Do

Generic openers:

  • "I hope this email finds you well"
  • "I'm reaching out because..."
  • "I wanted to introduce myself"

Feature dumps:

  • Long paragraphs about your product
  • Multiple value props at once
  • No clear CTA

Fake personalization:

  • "I noticed you work at [Company]" (obviously)
  • "Congrats on your role" (without context)

Markdown in emails:

  • Using bold or italic asterisks
  • Headers or formatted lists that won't render

Instead:

  • Lead with something specific you learned
  • One clear value prop
  • One clear ask
  • Plain text formatting only

Channel Selection

IF verified email available:
  → Email preferred (higher response rate)
  → Also provide LinkedIn backup

IF no email:
  → LinkedIn connection request
  → Follow-up message template for after connection

IF warm intro possible:
  → Suggest mutual connection outreach first

Company Configuration [CUSTOMIZE]

## Outreach Settings

- My name: [Your Name]
- My title: [Your Title]
- My company: [Company Name]
- Value prop: [One sentence - what you help with]

## Signature
[Your preferred email signature]

## Proof Points
- [Customer 1]: [Result]
- [Customer 2]: [Result]
- [Customer 3]: [Result]

## CTA Options
- Default: "Worth a 15-min call?"
- Softer: "Open to learning more?"
- Specific: "Can I send over a quick demo?"

## Tone
- [Professional / Casual / Direct]
- Industry-specific language: [Yes/No]

Example

Input: "draft outreach to the Head of Engineering at Notion"

Research finds:

  • Name: David Tibbitts
  • Recent news: Notion launched AI features
  • Hiring: 5 ML engineer roles open
  • Background: Ex-Facebook, Stanford CS

Output:

# Outreach Draft: David Tibbitts @ Notion

## Research Summary
**Target:** David Tibbitts, Head of Engineering at Notion
**Hook:** They're scaling AI features and hiring ML talent
**Goal:** Intro call about AI infrastructure

---

## Email Draft

**To:** david@notion.so
**Subject:** Notion's AI scaling + a thought

---

Hi David,

Saw Notion's AI rollout is gaining serious traction - congrats.
With 5 ML roles open, seems like you're scaling fast.

Curious how you're thinking about inference infrastructure
as usage grows. We helped [Similar Company] cut their AI
serving costs 40% while improving latency.

Worth a 15-min call to see if relevant to your roadmap?

Best,
[Name]

---

**Subject Alternatives:**
1. Notion AI + scaling question
2. Quick thought on Notion's ML hiring

---

## Email Draft Status
Draft created - check ~~email

來自 anthropic 的更多技能

access
anthropic
管理 Discord 頻道存取權限 — 核准配對、編輯允許清單、設定私訊/群組政策。當使用者要求配對、核准某人、查詢誰被允許時使用…
official
session-report
anthropic
從 ~/.claude/projects 的對話記錄中,生成一份可探索的 HTML 報告,內容涵蓋 Claude Code 工作階段的使用情況(包含 token、快取、子代理、技能及高成本提示)。
official
build-mcp-server
anthropic
當使用者要求「建立 MCP 伺服器」、「建立 MCP」、「製作 MCP 整合」、「為 Claude 包裝 API」、「暴露工具給…」時,應使用此技能。
official
cookbook-audit
anthropic
根據評分標準審核 Anthropic Cookbook 筆記本。每當要求進行筆記本審查或審核時使用。
official
handle-complaint
anthropic
處理進線客戶投訴的完整流程——提取背景資訊、草擬回覆,並建議營運改善方案。可接受選填的電子郵件或工單編號……
official
use-case-triage
anthropic
快速判斷某項處理活動是否需要PIA、強制性的GDPR DPIA,或可直接進行——揭露隱私政策衝突,並引導至正確的…
official
board-minutes
anthropic
根據您指定的格式,自動草擬董事會或委員會會議記錄。自動從您的日曆中偵測即將召開的董事會及委員會會議,詢問議程及…
official
renewal-tracker
anthropic
顯示即將到來的取消截止日期的合約,並在通知窗口關閉前發出警告,依據維護中的續約登記表進行操作。當用戶詢問時使用…
official