discovery-engine

name	discovery-engine
description	Find relevant people for outreach using Exa AI search capabilities. Supports LinkedIn search, company research, and cross-platform profile discovery. Use this skill when the user wants to find people based on topics, roles, companies, or industries.
allowed-tools	["Bash","Read","Write","Glob","Grep","TodoWrite","AskUserQuestion","WebSearch","WebFetch"]

Discovery Engine Skill

Finds relevant people for outreach using Exa AI integration.

When to Use This Skill

Use this skill when the user:

• Wants to find people to reach out to based on criteria
• Searches for people by role, company, industry, or location
• Needs to research companies before outreach
• Wants to build a target list for campaigns

When NOT to Use This Skill

Do NOT use this skill for:

• Sending messages → use workflow-engine or platform-specific skills
• Managing team → use team-manager
• Reading emails → use inbox-reader

Capabilities

1. People Search - Find people using Exa AI LinkedIn search
2. Company Research - Research companies for targeting
3. Profile Enrichment - Cross-reference to find Twitter/Instagram
4. Target Management - Store and organize discovered people
5. Export - Export to JSON/CSV for campaigns

MCP Selection Guidelines

CRITICAL: Select the correct MCP based on query requirements.

When to Use Exa MCP (exa-mcp-server)

Use for quick searches and exploration:

Scenario	Example
Quick lookup	"Find 5 marketers at Google"
Company info	"What does Acme Corp do?"
Small result set	< 10 prospects needed
Exploratory search	"Who are the AI startup founders?"

Exa Tools:

• linkedin_search_exa - Quick LinkedIn search
• company_research_exa - Company information
• web_search_exa - General web search

Characteristics:

• Fast response (seconds)
• Good for initial discovery
• Limited depth

When to Use Websets MCP (websets-mcp-server)

Use for deep research and exhaustive prospect lists:

Scenario	Example
Large prospect list	"Find 50 LinkedIn marketers"
B2B lead generation	"Build outreach list for AI startups"
Enriched data needed	"Find verified emails for founders"
Campaign-ready leads	"Prospect list for Series B CTOs"

Websets Tools:

• Exhaustive LinkedIn searches
• AI-powered data enrichment
• Verified contact information
• Bulk prospect discovery

Characteristics:

• Thorough, complete results
• Takes longer but more comprehensive
• Best for outreach campaigns

Auto-Selection Logic

def select_mcp(query: str, result_count: int) -> str:
    websets_keywords = ['prospects', 'leads', 'outreach', 'campaign',
                        'enriched', 'verified', 'comprehensive', 'list of']

    # Use Websets for large counts
    if result_count >= 10:
        return 'websets'

    # Use Websets for prospect-related keywords
    if any(kw in query.lower() for kw in websets_keywords):
        return 'websets'

    # Default to Exa for quick searches
    return 'exa'

MCP Configuration

Ensure MCPs are configured in .claude/.mcp.json:

{
  "exa": { "command": "npx", "args": ["-y", "exa-mcp-server"] },
  "websets": { "command": "npx", "args": ["-y", "websets-mcp-server"] }
}

Discovery Workflow

1. Create Discovery Session

Before searching, create a session to track results:

python .claude/scripts/discovery_engine.py session --query "AI startup founders San Francisco" --source linkedin_search

2. Perform Exa AI Search

Use Exa AI MCP to search (this happens through Claude's MCP integration):

Example prompt to Exa: "Search LinkedIn for AI startup founders in San Francisco who have raised Series A funding"

3. Process and Store Results

After receiving Exa results, add people to the session:

# The discovery_engine.py processes Exa responses
# Results are stored in output/discovery/people.json

4. View Discovered People

python .claude/scripts/discovery_engine.py list --has-linkedin

5. Export for Campaign

python .claude/scripts/discovery_engine.py export --format csv --output output/discovery/targets.csv

Commands Reference

View Discovery Stats

python .claude/scripts/discovery_engine.py stats

List Discovered People

# All people
python .claude/scripts/discovery_engine.py list

# Filter by criteria
python .claude/scripts/discovery_engine.py list --query "founder" --has-linkedin --status not_contacted

Get Person Details

python .claude/scripts/discovery_engine.py get --id PERSON_ID

Update Outreach Status

python .claude/scripts/discovery_engine.py status --id PERSON_ID --status contacted --notes "Sent LinkedIn connection"

Export People

# JSON export
python .claude/scripts/discovery_engine.py export --format json --output output/targets.json

# CSV export
python .claude/scripts/discovery_engine.py export --format csv --output output/targets.csv

Example Discovery Workflow

User: "Find 10 DevOps engineers at Series B startups in Austin"

Assistant Workflow:

1. Use TodoWrite to track:

   [
     {"content": "Create discovery session", "status": "in_progress"},
     {"content": "Search using Exa AI", "status": "pending"},
     {"content": "Process and store results", "status": "pending"},
     {"content": "Show discovered people", "status": "pending"}
   ]
   

2. Create session:

   python .claude/scripts/discovery_engine.py session --query "DevOps engineers Series B startups Austin" --source linkedin_search
   

3. Search with Exa AI: Use WebSearch or Exa MCP: "Search LinkedIn for DevOps engineers at Series B funded startups in Austin, Texas"

4. Process results: Parse Exa response and add people to discovery session.

5. Show summary:

   Found 10 DevOps engineers:
   
   1. John Smith - Senior DevOps at TechCo
      LinkedIn: linkedin.com/in/johnsmith
      Twitter: @john_devops
   
   2. Jane Doe - Platform Engineer at StartupX
      LinkedIn: linkedin.com/in/janedoe
      ...
   

6. Ask user: "Would you like to add these 10 people to a workflow for outreach?"

Data Storage

Discovery data is stored in:

• output/discovery/sessions.json - Search sessions
• output/discovery/people.json - All discovered people

Search Best Practices

1. Be Specific: "AI startup founders Series A San Francisco" > "startup founders"
2. Include Location: Helps narrow results
3. Add Context: Role + Industry + Company stage
4. Use Tags: Tag people for easy filtering later

Enrichment

After initial discovery, you can enrich profiles:

1. Find Twitter: Search web for "[Name] [Company] twitter"
2. Find Email: Look for public email on LinkedIn or company site
3. Company Research: Use company_research_exa for company context

The discovery engine automatically deduplicates and merges enriched data.