LinkedIn Job Signal →
Hiring Pain Point
Cold Email Machine

# Option A: Official one-liner (recommended)
curl -fsSL https://openclaw.ai/install.sh | bash

# Option B: npm global install
npm install -g openclaw@latest

# Verify installation
openclaw --version
# → openclaw v2026.2.23

openclaw onboard

# The wizard will prompt you for:
# ✓ LLM provider + API key (OpenAI / Anthropic / local Ollama)
# ✓ Gmail OAuth2 credentials
# ✓ Workspace name → "linkedin-sdr"
# ✓ Gateway port (default: 18789)
# ✓ Slack webhook URL

openclaw start
# Gateway running at http://localhost:18789
# Control UI available at http://localhost:18789/ui

# To run as a persistent background service:
openclaw start --daemon
# Or with PM2:
pm2 start openclaw -- start && pm2 save

# Install bankr skill from OpenClaw community skills library
openclaw skill install https://github.com/BankrBot/openclaw-skills/tree/main/bankr

# Install botchan skill (multi-agent routing helper)
openclaw skill install https://github.com/BankrBot/openclaw-skills/tree/main/botchan

# Verify installed skills
openclaw skill list
# → bankr@1.2.0  ✓
# → botchan@0.9.1 ✓

# 1. Create Google Cloud Pub/Sub topic
gcloud pubsub topics create openclaw-gmail-replies
gcloud pubsub subscriptions create openclaw-sub \
  --topic=openclaw-gmail-replies \
  --push-endpoint=http://YOUR_SERVER:18789/webhooks/gmail

# 2. Grant Gmail push permission
gcloud pubsub topics add-iam-policy-binding openclaw-gmail-replies \
  --member=serviceAccount:gmail-api-push@system.gserviceaccount.com \
  --role=roles/pubsub.publisher

# 3. Register in OpenClaw config (~/.openclaw/config.json)

{
  "workspace": "linkedin-sdr",
  "gateway_port": 18789,
  "channels": {
    "gmail": {
      "enabled": true,
      "pubsub_topic": "projects/YOUR_PROJECT/topics/openclaw-gmail-replies",
      "watch_labels": ["INBOX"],
      "sender_email": "outreach@yourdomain.com"
    },
    "slack": {
      "enabled": true,
      "webhook_url": "https://hooks.slack.com/services/YOUR/SLACK/WEBHOOK",
      "alert_channel": "#sdr-hot-replies"
    }
  },
  "tools": {
    "browser": { "enabled": true, "headless": true },
    "cron": { "enabled": true },
    "memory": { "enabled": true, "backend": "sqlite" },
    "shell": { "enabled": true, "whitelist": ["node", "curl", "python3"] }
  }
}

---
name: linkedin-signal-scraper
description: Monitors LinkedIn job postings for hiring signals every 4 hours
tools: [browser, cron, memory, shell]
cron: "0 */4 * * *"
---

## System Prompt

You are a B2B lead intelligence agent. Your job is to find companies
actively posting jobs that signal a pain point our product solves.

## Job Keywords to Monitor
Search LinkedIn Jobs for these exact title patterns (one search per run):
- "Head of Sales Operations" OR "RevOps Manager"
- "VP of Customer Success" OR "Customer Success Manager"
- "Sales Development Representative" (bulk hiring = scaling GTM)
- "Go-To-Market Lead"

For each job posting found:
1. Extract: company name, job title, company size, location, posting date, job URL
2. Check memory: skip if company already in contacted_leads list
3. Score signal strength (1-10) based on: seniority, urgency language, number of roles
4. If score ≥ 7: add to research_queue in memory
5. If score 4-6: add to warm_queue
6. Log results to shell: node log-leads.js

## Browser Instructions
Navigate to: https://www.linkedin.com/jobs/search/?keywords=QUERY&f_TPR=r86400
(r86400 = last 24 hours only — avoid re-processing old postings)
Extract all job cards on page 1. Do not login to LinkedIn.

---
name: lead-research-agent
description: Enriches leads from research_queue with Apollo + web research
tools: [browser, shell, memory]
trigger: memory_queue:research_queue
---

For each lead in research_queue:

## Step 1: Apollo.io Enrichment
Call Apollo API to find decision-maker contact:
  curl -X POST https://api.apollo.io/v1/people/search \
    -H "x-api-key: $APOLLO_API_KEY" \
    -d '{"q_organization_name": "{{company_name}}", 
         "person_titles": ["VP Sales", "CRO", "Head of Revenue", 
                           "CEO", "Founder", "VP Operations"]}'

Extract: first_name, last_name, email, linkedin_url, title, headline

## Step 2: Company News Research (browser)
Search: "{{company_name}} news site:techcrunch.com OR site:businesswire.com"
Look for: recent funding, product launches, new hires, expansions
Extract the most relevant news item from last 90 days.

## Step 3: Job Posting Deep Read (browser)
Navigate to the LinkedIn job URL. Extract:
- Full job description text
- "Nice to have" vs "Required" sections
- Tech stack mentioned
- Team size hints ("join our 5-person team")
- Pain language ("currently using spreadsheets", "manual process")

## Step 4: Compile Research Profile
Save to memory under lead_profiles:{{company_domain}}:
{
  "contact": {...apollo_data...},
  "company_news": "...",
  "job_signal": "...",
  "pain_points": [...],
  "personalization_hooks": [...],
  "signal_score": N
}

Move lead from research_queue → write_queue in memory.

---
name: sdr-copywriter-agent
description: Writes 3-touch hyper-personalized cold email sequences from lead profiles
tools: [memory]
trigger: memory_queue:write_queue
model: gpt-4o
---

You are an elite B2B cold email copywriter. Write emails that feel like
they were written by a thoughtful human who spent 20 minutes researching.
Never sound like a robot. Never use "I hope this email finds you well."

## Lead Context (injected from memory)
- Contact: {{contact.first_name}} {{contact.last_name}}, {{contact.title}}
- Company: {{company_name}}
- Job Signal: {{job_signal}}
- Recent News: {{company_news}}
- Pain Points: {{pain_points}}
- Personalization Hooks: {{personalization_hooks}}

## Email 1 — The Signal-Based Opener (Day 1)
Subject: max 8 words, reference the job posting or news item
Body: 
- Line 1: Hyper-specific observation about their job posting
- Line 2: Connect posting → implicit pain → your solution (1 sentence)
- Line 3: Social proof (metric or client name)
- CTA: Soft — "Worth a 15-min chat?" 
- Max 5 sentences total. NO fluff.

## Email 2 — The Value-Add Follow-up (Day 4)
Subject: Re: [Email 1 subject] OR new angle
Body:
- New insight or resource relevant to their pain
- One-sentence pitch
- Different CTA: Calendly link

## Email 3 — The Breakup (Day 10)
Subject: "Closing the loop, {{first_name}}"
Body: 2 sentences. Assume bad timing. Leave door open.
No pitch. Just human.

## Output Format
Return JSON with keys: email1_subject, email1_body, 
email2_subject, email2_body, email3_subject, email3_body

---
name: email-sender-agent
description: Sends email sequences via bankr skill with human-like timing
tools: [memory, cron]
skills: [bankr]
trigger: memory_queue:send_queue
---

For each lead in send_queue:

1. Retrieve email sequence from memory: lead_sequences:{{lead_id}}

2. Send Email 1 via bankr skill:
   bankr.send({
     to: "{{contact.email}}",
     from: "{{sender_name}} <outreach@yourdomain.com>",
     subject: "{{email1_subject}}",
     body: "{{email1_body}}",
     delay_seconds: random_between(30, 180)  // humanize sending
   })

3. Schedule follow-ups via cron:
   - Email 2: now + 4 days + random_hours(0, 8)
   - Email 3: now + 10 days + random_hours(0, 8)

4. Update memory: contacted_leads:{{company_domain}} = {
     status: "sequence_active",
     email1_sent: timestamp,
     thread_id: "{{gmail_thread_id}}"
   }

5. Log to shell for CRM sync.

Subject: Saw you're building out RevOps at Acme

Hi Sarah,

Noticed Acme just posted for a Head of Sales Operations — usually means 
pipeline visibility has become a genuine bottleneck at your growth stage.

We help Series B companies like yours get full-funnel reporting in place 
in under 2 weeks, without a 6-month RevOps hire — Rippling saved 
$180k in Year 1 using this approach.

Worth a 15-min chat this week?

[Your Name]

Subject: Re: Saw you're building out RevOps at Acme

Hi Sarah,

Sharing a quick 5-min read that might be timely — our breakdown of the
3 pipeline reporting mistakes most Series B teams make when scaling RevOps:
[Link]

If any of those land, I think we could save you 3+ months of the typical 
implementation timeline. Happy to show you a quick demo → 
https://cal.com/auralynxai/45min

[Your Name]

Subject: Closing the loop, Sarah

Hey Sarah — I'll assume the timing isn't right and won't bug you again.

If RevOps ever becomes a priority, you know where to find me.

[Your Name]

// Contact variables (from Apollo enrichment)
{{contact.first_name}}          // "Sarah"
{{contact.last_name}}           // "Chen"
{{contact.title}}               // "VP of Sales"
{{contact.linkedin_headline}}   // "Scaling GTM @ Acme"

// Company variables
{{company_name}}                // "Acme Corp"
{{company_size}}                // "51-200 employees"
{{company_funding_stage}}       // "Series B"

// Signal variables (extracted from job posting)
{{job_title_posted}}            // "Head of Sales Operations"
{{job_pain_language}}           // "manual processes", "spreadsheets"
{{job_tech_stack}}              // "Salesforce, HubSpot"
{{signal_urgency}}              // "immediate start", "ASAP"

// Research variables
{{company_recent_news}}         // "raised $12M Series B, Jan 2026"
{{personalization_hook}}        // Best hook chosen by research agent
{{pain_category}}               // "revops_broken" | "churn_risk" | etc.

---
name: reply-handler-agent
description: Classifies inbound replies and triggers appropriate actions
tools: [memory, shell]
channels: [gmail, slack]
trigger: channel:gmail:pubsub_push
---

On every Gmail Pub/Sub push event:

## 1. Match to Active Sequence
Check if sender email matches any lead in contacted_leads memory.
If no match: ignore (not our outreach thread).

## 2. Classify Reply Intent
Analyze reply body with LLM classification:

Classify into one of:
- POSITIVE: "yes", "interested", "let's chat", "send calendar"
- MAYBE: "not right now", "follow up next quarter", "send more info"  
- OBJECTION: "we already have X", "not our priority", price concern
- UNSUBSCRIBE: "remove me", "stop emailing", "unsubscribe"
- OUT_OF_OFFICE: auto-reply detected

## 3. Act on Classification

If POSITIVE:
  - Cancel pending follow-up crons for this lead
  - Send Slack alert to #sdr-hot-replies (within 60 seconds):
    "🔥 HOT REPLY from {{name}} @ {{company}}: '{{reply_snippet}}'"
  - Auto-reply with personalized email including Calendly link:
    "Hey {{first_name}}, great! Here's a link to grab time: 
     https://calendly.com/yourlink/30min — looking forward to it!"
  - Update memory: lead status → "meeting_requested"

If MAYBE:
  - Pause sequence (cancel Email 2/3 crons)
  - Schedule re-engagement cron for specified date + 7 days buffer
  - Send Slack alert to #sdr-warm-replies

If OBJECTION:
  - Send to Slack #sdr-objections for human review
  - Pause sequence pending human response

If UNSUBSCRIBE:
  - IMMEDIATELY cancel all crons for this lead
  - Add to do_not_contact list in memory
  - Log for compliance audit trail

If OUT_OF_OFFICE:
  - Extract return date if present
  - Reschedule Email 2 to: max(original_date, return_date + 2_days)

# SPF — authorize Google to send for your domain
TXT @ "v=spf1 include:_spf.google.com ~all"

# DKIM — add via Google Workspace Admin → Apps → Gmail → Authenticate email
# Google generates the key. Add the TXT record provided.

# DMARC — start with p=none to monitor, then move to p=quarantine
TXT _dmarc "v=DMARC1; p=none; rua=mailto:dmarc@yourdomain.com; pct=100"

# Custom tracking domain (optional but recommended for deliverability)
CNAME track "tracking.yourdomain.com"

---
name: deliverability-monitor
cron: "0 9 * * 1"  # every Monday 9am
---

Check Google Postmaster Tools API for each sending domain:
  - Spam rate: alert via Slack if > 0.1% (PAUSE if > 0.3%)
  - Domain reputation: alert if drops below HIGH
  - IP reputation: monitor for spamhaus listings

Alert format: "⚠️ DELIVERABILITY ALERT: {{domain}} spam rate {{rate}}% — 
investigate immediately before next send window."