Cold Email Icebreaker

The Cold Email Icebreaker agent generates highly personalized 7-email cold outreach sequences for a specific contact using deep research on the person and their company.

Overview

Agent ID: cold-email-icebreaker

Output: Self-contained HTML document with complete 7-email sequence, research summary, and personalization tags.

Execution Time: 30-60 seconds (extensive research phase)

Accepts: LinkedIn URL or email address

Sample Output

See it in action: Josh Aborwitz Email Sequence

This sample shows a complete 7-email sequence with person insights, company context, icebreaker hooks, and anti-AI detection guidelines.

Value Proposition

Creates cold email sequences that actually get responses by:

1. Deep personalization from contact research (posts, career history, interests)
2. Company-specific hooks (not generic value props)
3. Human-sounding copy that avoids AI tells
4. Full 7-email sequence with varied approaches

Quick Start

1. Start an Email Sequence Generation

curl -X POST "https://phoenix.hginsights.com/api/agents/v1/agents/cold-email-icebreaker/runs" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": {
      "contactIdentifier": "https://linkedin.com/in/username"
    },
    "params": {
      "campaignGoal": "meeting",
      "productContext": "Sales intelligence platform",
      "tonePreference": "conversational"
    }
  }'

Response:

{
  "run_id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "queued"
}

2. Poll for Completion

curl "https://phoenix.hginsights.com/api/agents/v1/runs/{run_id}" \
  -H "Authorization: Bearer YOUR_API_KEY"

3. Retrieve the HTML Sequence

curl "https://phoenix.hginsights.com/api/agents/v1/runs/{run_id}/artifacts" \
  -H "Authorization: Bearer YOUR_API_KEY"

Input Parameters

Required: Contact Identifier

Field	Type	Description
contactIdentifier	string	LinkedIn URL (e.g., https://linkedin.com/in/username) OR email address

Optional: Context Parameters

Field	Type	Default	Description
senderContext	object	{}	Info about sender (name, title, company, mutualConnections)
campaignGoal	string	meeting	Goal: meeting, reply, or referral
productContext	string	-	What you're selling (for relevance)
tonePreference	string	conversational	Tone: formal, conversational, bold, or casual

Example with Full Context

{
  "input": {
    "contactIdentifier": "https://linkedin.com/in/sarahchen-revops"
  },
  "params": {
    "senderContext": {
      "name": "Alex",
      "title": "Account Executive",
      "company": "HG Insights",
      "mutualConnections": ["Mike at Gong"]
    },
    "campaignGoal": "meeting",
    "productContext": "Sales intelligence data to improve pipeline quality",
    "tonePreference": "conversational"
  }
}

7-Email Sequence Structure

Email	Purpose	Day	Length
1	Icebreaker - Personal hook, establish relevance	1	50-80 words
2	Value - One specific pain point you solve	3	60-100 words
3	Social proof - Brief case study or name drop	6	50-80 words
4	Different angle - New hook or approach	10	60-100 words
5	Quick bump - Short, casual follow-up	14	30-50 words
6	Resource share - Offer something valuable, no ask	18	50-80 words
7	Breakup - Respectful close, leave door open	23	40-60 words

HTML Output

The agent produces a self-contained HTML document that includes:

1. Contact Header - Name, title, company, LinkedIn URL, email
2. Research Summary - Person insights and company insights
3. Icebreaker Hooks - Best personalization opportunities identified
4. 7-Email Cards - Each with:
   • Subject line
   • Email body
   • Personalization tags used
   • Word count and CTA type
5. Sequence Strategy - Primary hook, backup angles, best send times
6. Do Not Say List - AI tells specific to this prospect

Anti-AI Detection

The agent follows strict guidelines to avoid AI-sounding copy:

Banned Phrases

• "I hope this email finds you well"
• "I wanted to reach out"
• "I came across your profile/company"
• "Revolutionary/game-changing/cutting-edge"
• "Please let me know if you're interested"

Human-Sounding Patterns

• Start mid-thought sometimes
• Use "you" more than "I" or "we"
• One ask per email, max
• Vary sentence structure and length
• End with questions, not statements

Personalization Depth Levels

Level	Example	Quality
None	"Companies like yours..."	Bad
Surface	"I see you're at Acme"	Weak
Company	"With your EMEA expansion..."	Okay
Personal	"Your post about pipeline quality..."	Good
Specific	"When you said 'quality > quantity' last week..."	Best

Use Cases

AE Outbound Prospecting

Generate personalized sequences for high-value prospects before cold outreach.

BDR High-Volume Prospecting

Create varied, personalized emails that avoid the spam folder and AI detection.

Account-Based Marketing

Develop multi-touch sequences tailored to specific accounts and personas.

Re-engagement Campaigns

Create fresh approaches for prospects who've gone cold.

Error Handling

Scenario	Behavior
LinkedIn URL invalid	Returns HTML error page with format guidance
Contact not found	Proceeds with web search only, notes limited personalization
No personalization found	Flags as "low personalization risk," uses company-level hooks
Company unknown	Proceeds with contact-only personalization

MCP Tools Used

Tool	Purpose
contact_search	Find contact from LinkedIn URL or email
contact_enrich	Get full profile (career history, skills)
web_search	Find posts, talks, articles by the person
company_firmographic	Company context (size, industry)
company_technographic	Tech stack for relevance (optional)

Example: Python Integration

import requests
import time

API_KEY = "phx_your_api_key_here"
BASE_URL = "https://phoenix.hginsights.com/api/agents"

headers = {
    "Authorization": f"Bearer {API_KEY}",
    "Content-Type": "application/json"
}

def generate_cold_email_sequence(
    contact_identifier: str,
    product_context: str = None,
    campaign_goal: str = "meeting"
) -> str:
    """Generate a 7-email cold outreach sequence."""

    # 1. Start the run
    payload = {
        "input": {"contactIdentifier": contact_identifier},
        "params": {"campaignGoal": campaign_goal}
    }
    if product_context:
        payload["params"]["productContext"] = product_context

    response = requests.post(
        f"{BASE_URL}/v1/agents/cold-email-icebreaker/runs",
        headers=headers,
        json=payload
    )
    response.raise_for_status()
    run_id = response.json()["run_id"]
    print(f"Started run: {run_id}")

    # 2. Poll for completion
    while True:
        status_response = requests.get(
            f"{BASE_URL}/v1/runs/{run_id}",
            headers=headers
        )
        status = status_response.json()

        print(f"Status: {status['status']}")

        if status["status"] == "succeeded":
            break
        elif status["status"] == "failed":
            raise Exception(f"Run failed: {status.get('error')}")

        time.sleep(5)

    # 3. Get artifacts
    artifacts_response = requests.get(
        f"{BASE_URL}/v1/runs/{run_id}/artifacts",
        headers=headers
    )
    artifacts = artifacts_response.json()["artifacts"]

    html_artifact = next(a for a in artifacts if a["artifact_type"] == "html")
    return html_artifact["url"]

# Usage
sequence_url = generate_cold_email_sequence(
    "https://linkedin.com/in/username",
    "Sales intelligence platform"
)
print(f"Sequence available at: {sequence_url}")

Best Practices

1. Provide product context - Helps the agent tailor messaging to your specific solution
2. Include sender context - Mutual connections and sender info improve personalization
3. Review and customize - Use the sequence as a starting point, add your personal touch
4. Test different tones - Try bold for senior executives, casual for peers
5. Use the "Do Not Say" list - These phrases will trigger spam filters and AI detection

Next Steps

• Account Research Brief - Comprehensive account intelligence before outreach
• Buying Committee Mapper - Identify all stakeholders
• API Reference - Complete endpoint documentation