Research Lead

Add --post-to-slack to post the review card to Slack.

Execution Steps

Step 1: Scrape LinkedIn Profile

python3 .claude/skills/research-lead/scripts/scrape_linkedin.py "LINKEDIN_URL"

Output: JSON — name, headline, company, role, experience, recent posts, skills Dependencies: RELEVANCE_AI_API_KEY (for scraping API)

Step 2: Research Company (Perplexity)

python3 .claude/skills/research-lead/scripts/research_with_perplexity.py --company "Company Name" --person "Person Name" --role "Their Role"

Output: JSON — company overview, recent news, industry, challenges, tech stack signals Dependencies: PERPLEXITY_API_KEY

Step 3: AI Analysis (Parallel — runs 5 analyses simultaneously)

python3 .claude/skills/research-lead/scripts/analyze_with_openai.py --type lead_profile --profile profile.json --research research.json
python3 .claude/skills/research-lead/scripts/analyze_with_openai.py --type pain_gain_operational --profile profile.json --research research.json
python3 .claude/skills/research-lead/scripts/analyze_with_openai.py --type pain_gain_automation --profile profile.json --research research.json
python3 .claude/skills/research-lead/scripts/analyze_with_openai.py --type connection_request --profile profile.json --research research.json
python3 .claude/skills/research-lead/scripts/analyze_with_openai.py --type dm_sequence --profile profile.json --research research.json

Analysis types:

• lead_profile — Structured profile summary with relevance filter
• pain_gain_operational — Business pain points and potential gains
• pain_gain_automation — Automation-specific opportunities
• connection_request — Personalized LinkedIn connection message
• dm_sequence — 3-message DM sequence (value-first, not salesy)

Dependencies: OPENAI_API_KEY

Step 4: Generate Review Report

python3 .claude/skills/research-lead/scripts/generate_review_report.py --data combined_analysis.json

Output: HTML report for human review before sending anything

Step 5: Store in Google Sheets (Optional)

python3 .claude/skills/research-lead/scripts/update_google_sheet.py --data combined_analysis.json --sheet-id "SHEET_ID"

Step 6: Post to Slack (Optional)

python3 .claude/skills/research-lead/scripts/post_lead_review_to_slack.py --data combined_analysis.json

Process Flow

LinkedIn URL
   ↓
1. Scrape profile → profile.json
   ↓
2. Research company → research.json
   ↓
3. AI analysis (5x parallel) → analyses.json
   ↓
4. Generate review report → report.html
   ↓
5. (Optional) Store → Google Sheets
   ↓
6. (Optional) Post → Slack for review

Batch Processing

python3 .claude/skills/research-lead/scripts/batch_research_leads.py --source airtable

Pulls unprocessed leads from Airtable, runs the full pipeline for each.

Edge Cases

• Profile is private: Report what's available, note limitations
• Company not found: Research using available signals (headline, posts)
• Rate limits: Built-in retry with exponential backoff in each script
• Missing API keys: Scripts fail gracefully with clear error message about which key is missing

Environment Variables Required

RELEVANCE_AI_API_KEY=     # LinkedIn scraping
PERPLEXITY_API_KEY=       # Company research
OPENAI_API_KEY=           # AI analysis
SLACK_BOT_TOKEN=          # Optional: Slack posting
GOOGLE_SHEETS_CREDENTIALS= # Optional: Google Sheets storage

Cost

~$0.40 per lead (API calls combined). Pipeline takes 45-60 seconds.