Data Enricher

# SKILL.md - Data Enricher

## Purpose
Enrich leads with email addresses and format data for Notion.

## Model to Use
- **ollama/llama3.2:8b** (FREE) for data formatting
- **haiku** for Hunter.io API calls

## Rate Limits
- Max 10 Hunter.io lookups per session (API limit)
- 5 seconds between API calls
- Batch similar domains together

## Email Discovery Methods (In Order)

### 1. Website Contact Page
- Check /contact, /about, /pages/contact
- Look for mailto: links
- Check footer

### 2. Instagram Bio
- Check bio for email
- Check "Contact" button

### 3. Hunter.io API
```
GET https://api.hunter.io/v2/domain-search
?domain={domain}
&api_key={HUNTER_API_KEY}
```

Response includes:
- emails[]
- confidence score
- type (generic/personal)

**Only use emails with confidence > 70%**

### 4. Email Pattern Guessing
Common patterns:
- hello@domain.com
- info@domain.com
- contact@domain.com
- [firstname]@domain.com

## Email Priority
1. Founder/owner personal email (best)
2. hello@ or hi@ (good)
3. info@ or contact@ (okay)
4. Generic support@ (last resort)

## Output Format
```
{
  "domain_key": "brandname.com",
  "brand_name": "Brand Name",
  "niche": "skincare",
  "website_url": "https://brandname.com",
  "ig_handle": "@brandname",
  "followers_est": 15000,
  "contact_email": "hello@brandname.com",
  "email_confidence": "high",
  "email_source": "hunter.io",
  "source": "meta_ads",
  "status": "new"
}
```

## Deduplication
Before adding any lead:
1. Normalize domain: lowercase, remove www., remove https://
2. Check if domain_key exists in Notion
3. If exists, skip (don't duplicate)
4. Log: "Skipped [domain] - already in pipeline"

## Batch Processing
- Process 10 leads at a time
- Format all data before Notion sync
- Save formatted batch to workspace/leads-enriched-YYYY-MM-DD.json