什么是 冷启动获客猎手？

冷启动获客猎手（Cold Outreach Hunter）是为利用 Openclaw Skills 构建高性能外向型销售引擎的开发人员和增长团队设计的复杂元技能。它作为主编排器，协调包括用于线索发现的 Apollo、用于个人资料背景的 LinkedIn 以及用于高转化文案策略的 YC 冷启动框架在内的专业工具。通过将这些分散的服务整合到单个统一的工作流程中，它消除了通常与在潜在客户开发和消息平台之间移动数据相关的各种手动开销。

此技能专为在 B2B 沟通中优先考虑质量和相关性的人员而构建。它强制执行严格的质量关卡，确保生成的每个序列都基于可验证的潜在客户数据，并遵循成熟的初创公司增长方法论。无论您是发布新服务还是扩展现有的销售动作，此技能都能在 Openclaw Skills 框架内提供专业、非推销感且高度自动化的潜在客户开发方法。

下载入口:https://github.com/openclaw/skills/tree/main/skills/h4gen/cold-outreach-skill

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。

冷启动获客猎手 应用场景

• 针对特定的理想客户画像（ICP），如 SaaS CMO 或工程副总裁，自动化线索搜寻。
• 基于 YC 风格的简洁和转化原则，生成个性化的多步骤电子邮件序列。
• 利用 LinkedIn 背景丰富潜在客户数据，创建非模版化的、高度相关的破冰语。
• 为设计服务、技术审计或高价值会议请求开发准备就绪的 B2B 活动草案。
• 在保持以人为本的语调并严格遵守外发政策的同时，扩展外向型业务规模。

1. 线索发现：该技能利用 Apollo API，根据您特定的 ICP 过滤器和地理位置限制，构建一个高度匹配的潜在客户库。
2. 数据丰富：它尝试通过 LinkedIn 和 Apollo 收集更深层的背景信息，确保每个潜在客户都有经过验证的电子邮件和清晰的职业档案。
3. 策略制定：使用 YC 冷启动框架，它为每个潜在客户创建策略简报，重点关注特定问题、解决方案和证明点。
4. 序列生成：它与 MachFive（冷邮件）对接，生成包含主题和正文的结构化、多步骤电子邮件序列。
5. 质量保证与交付：该技能会对个性化的准确性进行最终验证检查，并提供包含调度建议的待发送方案包。

冷启动获客猎手 配置指南

要使用此元技能，请确保您的 Openclaw Skills 环境中配置了所需的依赖项和 API 密钥。

# 安装所需的技能依赖项
npx -y clawhub@latest install apollo-api linkedin-api yc-cold-outreach cold-email

# 配置 API 凭据
export MATON_API_KEY="your_maton_key"
export MACHFIVE_API_KEY="your_machfive_key"

冷启动获客猎手 数据架构与分类体系

该技能将外发数据组织成结构化的输出包，以便轻松交付给发送系统：

name: cold-outreach-hunter
description: Meta-skill for orchestrating Apollo API, LinkedIn API, YC Cold Outreach, and MachFive Cold Email into a complete B2B cold outreach pipeline. Use when the user wants end-to-end lead sourcing, enrichment, personalized copy strategy, and generation-ready outreach sequences with strict quality and safety gates.
homepage: https://clawhub.ai
user-invocable: true
disable-model-invocation: false
metadata: {"openclaw":{"emoji":"dart","requires":{"bins":["python3","npx"],"env":["MATON_API_KEY","MACHFIVE_API_KEY"],"config":[]},"note":"Requires local installation of apollo-api, linkedin-api, yc-cold-outreach, and cold-email."}}

Purpose

Run a full B2B cold outreach workflow from ICP definition to sequence-ready output.

Primary objective:

• Identify high-fit leads.
• Enrich context for personalization.
• Produce concise, non-salesy, high-response outreach sequences.
• Return execution-ready assets for external sending/scheduling systems.

This is an orchestration skill. It coordinates upstream skills; it does not replace them.

Required Installed Skills

• apollo-api (inspected latest: 1.0.5)
• linkedin-api (inspected latest: 1.0.2)
• yc-cold-outreach (inspected latest: 1.0.1)
• cold-email (MachFive Cold Email, inspected latest: 1.0.5)

Install/update with ClawHub:

npx -y clawhub@latest install apollo-api
npx -y clawhub@latest install linkedin-api
npx -y clawhub@latest install yc-cold-outreach
npx -y clawhub@latest install cold-email
npx -y clawhub@latest update --all

Verify availability:

npx -y clawhub@latest list

If any required skill is missing, stop and report exact install commands.

Required Credentials

• MATON_API_KEY for apollo-api and linkedin-api (Maton gateway)
• MACHFIVE_API_KEY for cold-email

Preflight checks:

echo "$MATON_API_KEY" | wc -c
echo "$MACHFIVE_API_KEY" | wc -c

If either key is missing or empty, stop before lead processing.

Job Context Template

Collect these inputs before execution:

• offer: what is being sold (example: design service)
• icp_title: target role (example: CMO)
• icp_industry: target industry (example: SaaS)
• icp_location: target location (example: Berlin)
• lead_count_target (example: 50)
• campaign_goal: reply, meeting, referral, audit request, etc.
• proof_points: case studies, metrics, social proof
• tone_constraints: plain-English, short, non-salesy
• machfive_campaign (campaign ID or campaign name to resolve)
• execution_mode: draft-only or generation-ready

Do not start writing copy until these are explicit.

Tool Responsibilities

Apollo API (apollo-api)

Use for lead discovery and basic enrichment.

Operationally relevant behavior from inspected skill:

• Search people: POST /apollo/v1/mixed_people/api_search
• Search filters include:
  • q_person_title
  • person_locations
  • q_organization_name
  • q_keywords
• Enrich person by email or LinkedIn URL:
  • POST /apollo/v1/people/match
• Supports pagination via page and per_page.
• Uses Maton gateway and optional Maton-Connection header.

Primary output of this stage:

• initial lead list with role/company/email/linkedin_url (when available)

LinkedIn API (linkedin-api)

Use for LinkedIn-side context where accessible through provided endpoints.

Operationally relevant behavior from inspected skill:

• Authenticated profile/user info endpoints (for connected account context).
• Content/posting APIs (ugcPosts) and organization post/stat APIs.
• Requires MATON_API_KEY and LinkedIn protocol headers.

Important boundary:

• The inspected skill is not a generic scraper for arbitrary third-party personal profiles and recent personal posts.
• If a workflow requires deep per-lead personal-post enrichment, mark that as additional-tool-required.

YC Cold Outreach (yc-cold-outreach)

Use as writing strategy/critique framework, not as a transport API.

Core principles to enforce:

• single goal per email
• human tone
• deep personalization (not just token replacement)
• brevity/mobile readability
• credibility and proof
• reader-centric language
• clear CTA

MachFive Cold Email (cold-email)

Use for sequence generation from prepared lead records.

Operationally relevant behavior from inspected skill:

• Campaign required (campaign_id mandatory for generate endpoints).
• Single lead sync generation (/generate) can take minutes; use long timeout.
• Batch async generation (/generate-batch) returns list_id; poll list status; export when complete.
• Lead email is required.
• Supports structured sequence output with subject/body per step.

Canonical Workflow

Stage 1: Build lead universe (Apollo)

1. Query Apollo for ICP-constrained leads (example: CMO + SaaS + Berlin).
2. Page until lead_count_target or quality threshold is reached.
3. Normalize each lead record to required fields.
4. Drop records without email if generation-ready mode is requested (MachFive requires email).

Recommended normalized lead schema:

{
  "lead_id": "apollo-or-derived-id",
  "name": "Anna Example",
  "title": "Chief Marketing Officer",
  "company": "Startup GmbH",
  "location": "Berlin",
  "email": "anna@startup.com",
  "linkedin_url": "https://linkedin.com/in/...",
  "source": "apollo-api"
}

Stage 2: Enrich personalization context

1. Attempt LinkedIn/API enrichment within supported endpoints.
2. If direct personal-post signal is unavailable, keep the context slot explicit as not_available.
3. Optionally enrich from Apollo fields (company, role, keywords, domain context) to avoid fake personalization.

Personalization object per lead:

{
  "icebreaker": "not_available_or_verified_fact",
  "pain_hypothesis": "Likely CRO bottleneck in paid landing pages",
  "proof_hook": "Helped X improve conversion by Y%",
  "confidence": 0.0
}

Hard rule:

• Never invent a post, interest, or quote.

Stage 3: Message strategy (YC framework)

For each lead, create a strategy brief before generating copy:

• Problem: what specific pain this role likely has
• Solution: what your offer solves
• Proof: one concrete metric/client signal
• CTA: one low-friction next step

Apply YC constraints:

• one ask
• short/mobile-first
• human language
• personalization grounded in verifiable context

Stage 4: Sequence generation (MachFive)

1. Resolve campaign ID first (GET /api/v1/campaigns) if not provided.
2. Submit leads with required email field.
3. Prefer batch for many leads; poll until completion.
4. Export JSON result and map sequences back to lead IDs.

Required generation payload hygiene:

• include name, title, company, email
• include linkedin_url and company_website when available
• set email_count intentionally (usually 3)
• use approved CTA set aligned with campaign goal

Stage 5: QA and decision gate

Before declaring output ready, validate each sequence:

• personalization factuality check
• YC rubric check (human, concise, one CTA)
• token insertion sanity (name/company/title correct)
• prohibited claims check (no fabricated proof)

Any failed sequence must be flagged needs_revision.

Stage 6: Scheduling and send handoff

This meta-skill outputs send-ready recommendations, not direct send automation.

If user asks for timing optimization (for example Tuesday 10:00), return it as a scheduling recommendation field and handoff plan.

Example handoff object:

{
  "lead_id": "...",
  "sequence_status": "approved",
  "suggested_send_time_local": "Tuesday 10:00",
  "timezone": "Europe/Berlin",
  "send_system": "external",
  "notes": "Timing is recommendation-only; execution tool must schedule/send."
}

Causal Chain (Scenario Mapping)

For the scenario "sell design services to startup marketing leaders":

1. Apollo returns target leads (example target: 50 CMOs in Berlin SaaS).
2. LinkedIn/API enrichment attempts to add usable context per lead.
3. YC framework converts lead context into a concise Problem → Solution → Proof → CTA angle.
4. MachFive generates multi-step sequences with validated variables.
5. Agent outputs:
   • approved sequences
   • quality score per lead
   • scheduling recommendation (example: Tuesday 10:00 local)

Output Contract

Always return these sections:

• LeadSummary

  • requested vs qualified lead count
  • rejection reasons (missing email, poor fit, duplicate)

• EnrichmentSummary

  • fields successfully enriched
  • unavailable fields and why

• SequencePackage

  • one object per lead with subjects/bodies by step
  • QA status (approved or needs_revision)

• ExecutionPlan

  • send-time recommendation
  • required external sender/scheduler
  • blockers (missing campaign, missing API key, missing email)

Guardrails

• Never fabricate personalization facts.
• Never claim a lead posted something unless sourced and verifiable.
• Do not proceed to MachFive generation without campaign ID resolution.
• Do not mark sequence approved when CTA is unclear or multiple asks exist.
• Keep language non-manipulative and compliant with outreach policies.

Failure Handling

• Missing MATON_API_KEY: stop Apollo/LinkedIn stages.
• Missing MACHFIVE_API_KEY: stop generation stage and return draft-only strategy.
• Missing campaign ID: list campaigns and request explicit selection.
• Batch timeout/partial output: continue via list status + export recovery flow.
• Insufficient lead quality: return reduced high-quality set instead of forcing volume.

Known Limits from Inspected Upstream Skills

• linkedin-api inspected capability set is not equivalent to unrestricted scraping of arbitrary personal lead activity.
• cold-email generates sequences but does not itself guarantee outbound send scheduling/execution.
• apollo-api provides search/enrichment primitives; email deliverability validation beyond provider fields may require extra tooling.

Treat these as explicit constraints in planning and reporting.