Token导航 LogoToken导航

Agent Skill · 研究检索

email-drafting(电子邮件起草)

email-drafting 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

email-drafting 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 68,166。

总安装 235周安装 10GitHub Stars 607下载 82核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:email-drafting(电子邮件起草)
来源仓库:https://github.com/athina-ai/goose-skills
仓库路径:skills/email-drafting
安装命令:
npx skills add https://github.com/athina-ai/goose-skills --skill email-drafting
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/athina-ai/goose-skills --skill email-drafting

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

基于真实成功案例撰写高回复率冷电子邮件草稿。提供框架而非固定脚本,强调个性化与情境适配能力。支持主题行、正文与序列设计一体化输出。

安装命令:npx skills add https://github.com/athina-ai/goose-skills --skill email-drafting 需明确产品类型、目标角色与沟通目标以定制话术。

来源仓库内路径:skills/email-drafting

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/athina-ai/goose-skills --skill email-drafting;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

email-drafting 怎么安装?

在终端执行 npx skills add https://github.com/athina-ai/goose-skills --skill email-drafting。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

email-drafting 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

email-drafting 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

email-drafting 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

email-drafting 用的人多吗?

公开数据是累计安装 235 次,近一周安装 10 次,下载 82 次,GitHub 607 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

email-drafting 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
email-drafting

电子邮件起草 · 本页

235607需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Email Drafting

Pure reasoning skill for writing cold emails. No scripts, no tools — just frameworks, patterns, and examples from real campaigns that consistently generate replies.

When to Use

Use this skill when:

  • User says "write a cold email", "draft outreach", "help me with email copy", "write a sequence"
  • Any task requires cold email copy — subject lines, full sequences, or individual emails

Phase 0: Intake

Collect campaign context before writing anything. Ask all questions at once, organized by category. Skip any the user has already answered.

Campaign Context

  1. What product/service are you selling?
  2. What problem does it solve? Who feels this pain most acutely?
  3. What's the campaign angle? (hiring signal, competitor displacement, pain-based, event-triggered, etc.)
  4. Is there a specific signal or trigger? (job posting, G2 review, LinkedIn engagement, funding round, etc.)

Audience

  1. Who is the recipient? (title, seniority, department)
  2. What keeps them up at night? (daily frustrations relevant to your product)
  3. What objections will they have? (budget, switching cost, "we already have X", timing)

Proof & Credibility

  1. What social proof do you have? (customer logos, case studies, metrics)
  2. Name 2-3 peer companies the recipient would recognize as similar to them
  3. Any hard metrics? (cost savings, speed improvement, % lift)

Tone & Style

  1. What tone fits? (casual-direct, professional-sharp, provocative, empathetic)
  2. Who is the sender? (founder, AE, SDR — this affects voice)
  3. Any brand guidelines or words to avoid?

Sequence

  1. How many touches? (default: 3)
  2. What's the desired CTA? (call, demo, reply, resource download)
  3. Email-only or multi-channel? (email + LinkedIn, email + phone)

Phase 1: Draft Emails

Email Structure Formula

Every cold email follows this skeleton:

Hook (1 sentence) → Evidence (1-2 sentences) → Offer (1 sentence)

Word count targets:

  • Cold intro (Touch 1): 50-90 words
  • Follow-up (Touch 2-3): 30-50 words
  • Breakup (final touch): 20-40 words

Frameworks

Pick the framework that matches the campaign angle:

FrameworkStructureBest For
PASProblem → Agitate → SolvePain-based signals (complaint posts, operational friction)
BABBefore → After → BridgeAspirational buyers (growth-stage, scaling companies)
AIDAAttention → Interest → Desire → ActionCold database outreach (no specific signal)
Signal-Proof-AskSignal → Proof → Soft askSignal-based campaigns (hiring, engagement, events)

Personalization Tiers

Choose based on campaign size and expected ROI per lead:

TierWhat It MeansLead VolumeExpected Reply Rate
Tier 1 (Generic)Merge fields only ({first_name}, {company}). Same template for everyone.500+ leads1-3%
Tier 2 (Segment)Industry/role-specific pain points + proof swaps. One template per segment.50-500 leads3-7%
Tier 3 (Deep)Reference a specific signal (their post, comment, job posting, news). Unique per lead.1-50 leads8-20%

Subject Line Patterns

8 proven patterns from real campaigns:

#PatternExample
1Signal reference"Before you fill that [role] role"
2Peer framing"What TIA members are doing with AI workers"
3Question"Is [Company] still [doing thing product fixes]?"
4Replacement"Looking for a [competitor] replacement?"
5Data hook"$150K agency study → $8K. 48 hours."
6Empathy"When [event that affected them]"
7Direct"[Topic] for [Company]"
8Curiosity"How [peer company] did [interesting thing]"

Rules for subject lines:

  • Under 50 characters
  • No ALL CAPS, no exclamation marks, no emoji
  • No "quick question" or "touching base"

Tone Guidance

ToneWhen to UseVoice Example
Casual-DirectSDR sending to peers, startup-to-startup"Hey — saw your post. We work on the same problem."
Professional-SharpEnterprise outreach, VP+ recipients"I wanted to reach out because [specific reason]."
ProvocativeCompetitive displacement, challenger positioning"Your current tool is costing you more than you think."
EmpatheticOrphan capture, pain-based outreach"I know switching platforms mid-cycle is brutal."

Sequence Design Principles

TouchTimingPurposeLengthNotes
Touch 1Day 1Hook + proof + soft CTA50-90 wordsThe only email that can be longer
Touch 2Day 3-5New angle or asset30-50 wordsDifferent proof point, not a "bump"
Touch 3Day 7-10Different proof point or social20-40 wordsShorter = better this late
Touch 4Day 14-21Breakup (optional)20-30 wordsRemove pressure, leave door open

Sequence rules:

  • Never repeat the same CTA across touches
  • Each touch needs a new reason to reply
  • Later touches = shorter emails
  • Never send a "just checking in" or "bumping this" — add value or stop

Hard Rules

These are non-negotiable. Every email must pass all 10:

  1. No filler openers. Never "I hope this finds you well", "I hope you're having a great week", "just reaching out"
  2. No "just checking in" follow-ups. Every touch adds a new reason to reply
  3. Max 4 paragraphs per email. Most should be 2-3
  4. Every email references something specific to the recipient. Title, company, signal, industry — never fully generic
  5. Exactly one CTA per email. Always low-friction (15-min call, "worth a look?", "open to chatting?")
  6. Never lie about how you found them. If it was a database search, don't say "I came across your profile"
  7. No filler words. Ban: synergy, leverage, circle back, loop in, touch base, align, ping
  8. Subject lines under 50 chars. No caps, no exclamation marks, no emoji
  9. No selling in the first sentence. Lead with them, not you
  10. Sign off simply. Name only, or Name + one-line title. No "Best regards", no "Looking forward to hearing from you"

Phase 2: Review & Refine

  1. Present 3-5 draft variants for Touch 1 (different angles/frameworks)
  2. Ask user to pick a direction or combine elements
  3. Generate the full sequence based on chosen direction
  4. Iterate max 3 rounds — after that, ship it

Example Library

Real emails from real campaigns. Use these as structural templates — swap product, proof, and signal for the current campaign.

Type 1: Signal-Based (Hiring)

Signal source: LinkedIn Jobs — company posting for role that product replaces Framework: Signal-Proof-Ask

Subject: Before you fill that {role_title} role

Hi {first_name} — I noticed you're hiring for a {role_title} at {company}. Before you finalize that hire, worth a quick look at what companies like DHL, Werner, and MODE Global are doing instead — deploying AI workers for exactly this function. 4x cheaper than a BPO equivalent. 10x the call capacity. Runs 24/7. Happy to send over a quick overview or jump on a 15-minute call if the timing is right.

Why it works: Opens with their specific hiring signal (proves relevance). Names peer companies (social proof). Quantifies the alternative (4x, 10x). Low-friction CTA.


Type 2: Signal-Based (Peer/Association)

Signal source: Industry association membership directory Framework: BAB (Before → After → Bridge)

Subject: What TIA members are doing with AI workers

Hi {first_name} — as a TIA member running a serious freight brokerage, you're probably seeing the same thing we are: the best operators are rethinking how they staff carrier-facing functions. Companies like Werner, MODE Global, and Circle Logistics — TIA members you'd recognize — have already deployed AI workers for carrier calls, check calls, and load coordination. I wanted to reach out because HappyRobot works specifically with freight brokers at this scale. Worth a 15-minute call to see if it's relevant for {company}?

Why it works: Peer framing ("TIA members you'd recognize") creates belonging pressure. Names real members using the product. Positions as industry-specific, not generic.


Type 3a: Signal-Based (Competitor Engagement — Liker)

Signal source: LinkedIn Sales Navigator — engaged with competitor company page content Framework: Signal-Proof-Ask

Subject: The other side of freight AI

Hi {first_name} — I noticed you've been following the freight AI space and wanted to reach out. HappyRobot is the other side of that conversation — we're the platform 8 of the top 10 US freight brokers including DHL and Werner run for AI carrier calls. If you're evaluating options in this space, I'd love to show you what enterprise-grade looks like. 15 minutes?

Why it works: Acknowledges their research without calling out the competitor by name in email (saves that for LinkedIn). "Other side of that conversation" is intriguing. Proof point (8 of top 10) is specific and strong.


Type 3b: Signal-Based (Competitor Engagement — Commenter)

Signal source: Same as 3a but lead left a comment (higher intent) Framework: Signal-Proof-Ask with comment reference

Subject: Re: your comment on freight AI

Hi {first_name} — I saw your comment on {competitor}'s post about {topic}. That's exactly the problem we work on. HappyRobot powers AI carrier calls for DHL, Werner, and 8 of the top 10 US freight brokers. If you're actively looking at this, worth 15 minutes to see the other option?

Why it works: References their exact comment (Tier 3 personalization). "Actively looking at this" matches their behavior. Even shorter than 3a because higher-intent leads need less convincing.


Type 4: Pain-Based (Commenter)

Signal source: LinkedIn post search for pain-language keywords Framework: PAS (Problem → Agitate → Solve)

For commenters (wrote about the pain):

Hi {first_name}, saw your comment on the {source} post about {pain_topic}. "{comment_snippet}..." — that resonated. We've been working with brokers who had the same issue. HappyRobot handles {relevant_task} so your team doesn't have to. Worth a quick look?

For reactors (liked/reacted to pain content):

Hi {first_name}, noticed you've been following the conversation around {pain_topic} on LinkedIn. If {company} is dealing with {specific_pain}, we might be able to help. HappyRobot automates {relevant_task} for freight brokers — no extra headcount needed. Open to a quick chat?

Why it works: Quotes their own words back to them (strongest personalization). Empathetic tone — no hard sell. Short. The reactor version is lighter-touch because the signal is weaker.


Type 5: Competitive Displacement

Signal source: G2 reviews (1-star), public complaints, known competitor customers Framework: PAS

Subject: UserTesting alternatives in 2026

{first_name} — I'll keep this short. If you've run into participant quality issues with UserTesting, you're not alone. HubSpot and Glassdoor both moved to Outset after hitting the same wall. We use Prolific-verified participants instead of a general panel. Different quality tier entirely. Worth 20 minutes to compare?

Why it works: Validates their frustration without bashing the competitor. Names companies they'd respect. Explains the "why" behind the switch (Prolific-verified participants). CTA is specific (20 minutes, comparison framing).


Type 6: ABM / Enterprise

Signal source: Target account list, vertical-specific research Framework: AIDA

Touch 1:

{first_name} — WeightWatchers cut their per-study research cost from $150K to $8K using Outset's AI-moderated interviews. They run 3x the studies they used to, in half the time. I put together a quick breakdown of what this could look like for {company}'s {vertical} research. Worth 20 minutes?

Touch 3 (Day 7):

Following up with something concrete — built a rough ROI model using {industry} benchmarks. Even conservative numbers show a 6-8x reduction in per-study cost. Happy to walk through the math. 15 minutes?

Touch 5 (Day 14):

Last note from me. If research cost or speed isn't a priority right now, totally understand. But if it is — the WeightWatchers and HubSpot numbers are worth seeing. Open to a quick benchmark call? No demo, just data.

Why it works: Leads with a specific customer story and hard numbers. Each touch adds new value (case study → ROI model → benchmark offer). Breakup touch removes pressure. CTA evolves: overview → walk-through → benchmark.


Type 7: Follow-Up Patterns (Touch 2+)

New angle (not a bump):

{first_name} — different angle from my last email. {Peer company} just published results from their first quarter using {product}: {specific metric}. Thought it might be relevant given {company}'s {situation}.

Asset-led:

Put together a one-page breakdown of {topic relevant to them}. No pitch — just data. Want me to send it over?

Social proof drop:

Quick update — {new customer} just went live with us last week. Similar setup to {company}. Happy to share what their onboarding looked like.

Breakup:

I'll keep this short — if the timing isn't right, no worries at all. But if {problem} comes back up, I'm an easy call away. Cheers.

Type 8: Re-engagement (90-Day)

Signal source: Outreach log — leads contacted 90+ days ago with no reply, or leads who filled the role your product replaces Framework: Signal-Proof-Ask

Hi {first_name} — we chatted about 3 months ago when you were hiring for {role}. Curious how it's going. If you're still feeling the pain on {problem}, a few things have changed on our end worth seeing. 15 minutes to catch up?

Why it works: References the original conversation and signal. Acknowledges time has passed. "A few things have changed" creates curiosity without overselling.

Output Format

When delivering email drafts, use this structure:

Single Email

**Subject:** [subject line]
**Personalization tier:** [1/2/3]
**Framework:** [PAS/BAB/AIDA/Signal-Proof-Ask]
**Word count:** [X words]

---

[Email body with merge fields in {curly_braces}]

---

**Merge fields used:** {first_name}, {company}, {role_title}, ...

Full Sequence

## Sequence: [Campaign Name]

**Touches:** [N]
**Personalization tier:** [1/2/3]
**Tone:** [casual-direct / professional-sharp / provocative / empathetic]

### Touch 1 — Day 1
**Subject:** [subject]
**Framework:** [framework]

[body]

### Touch 2 — Day [N]
**Subject:** Re: [original subject] OR [new subject]
**Framework:** [framework]

[body]

### Touch 3 — Day [N]
**Subject:** [subject]

[body]

---

**Merge fields:** {first_name}, {company}, ...
**Notes:** [any special instructions for the outreach tool]

平台分布

Codex

33.13%
按下载量换算27

Claude

31.32%
按下载量换算26

Cursor

19.4%
按下载量换算16

Gemini CLI

9.13%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills