leaddelta-intelligence

LinkedIn relationship intelligence via LeadDelta MCP. Triggers when user asks to 'find warm intro', 'search LinkedIn connections', 'who knows [person/company]', 'find a connector', 'LinkedIn path', or

SKILL.md

GitHub

LeadDelta Relationship Intelligence Skill

LinkedIn-native relationship intelligence combining LeadDelta's Warm Intro AI (10+ signals) with Nexus Neptune multi-hop graph traversal.

Prerequisites

  • Read _shared/plugin-context.md for Nexus domain context
  • LeadDelta connection required: API key must be configured in Nexus admin (/admin/ai-config → Connectors tab)
  • Falls back to Neptune-only paths if LeadDelta is unavailable (surfaces this to user)
  • MCP tools available: leaddelta_search_connections, leaddelta_warm_intro_ai, all nexus_* relationship tools

When This Skill Triggers

User Intent Example Phrases
Warm intro finding "Who knows someone at Blackstone?", "Find a warm intro to KKR"
LinkedIn search "Search LinkedIn for", "Who in our network works at...", "Find connections at..."
Connector discovery "Who can introduce me to...", "Best person to reach out through"
Relationship intelligence "LinkedIn relationship path", "How connected are we to..."

Protocol Overview

Intent Classification → LinkedIn Search → Neptune Paths → LeadDelta Warm Intro AI → Combined Scoring → Draft Message
       (parallel)

Phase 1: Intent Classification

Classify the request:

  • Company intro: Find paths to any person at a target company
  • Person intro: Find paths to a specific named individual
  • Network search: Discover relevant people matching criteria
  • Path mapping: Map connections between two specific people

Phase 2: Dual-Network Search

Run searches in parallel:

LeadDelta LinkedIn Network:

leaddelta_search_connections({
  query: "[natural language from user]",
  filters: { company, title, location, industry },
  limit: 25
})

Returns: LinkedIn connections with stage, tags, last interaction, connection date.

Nexus Internal Network:

nexus_search_network({
  query: "[same natural language query]",
  company, title, connection_strength,
  limit: 20
})

Returns: CRM contacts with relationship metadata.

Phase 3: Warm Intro Path Discovery

LeadDelta Warm Intro AI (when target company/person known):

leaddelta_warm_intro_ai({
  target_company: "...",
  target_person: "...",      // optional
  min_intro_score: 0.5,
  max_results: 10
})

LeadDelta scores 10+ signals per potential connector:

  1. Length of overlapping employment (weighted highest)
  2. Recency of LinkedIn connection
  3. LinkedIn messaging activity history
  4. Educational overlap (institution + graduation proximity)
  5. Department/role proximity at shared employer
  6. Mutual connections count
  7. Geographic proximity
  8. Profile view / engagement history
  9. Response rate signals
  10. Proprietary LeadDelta relationship indicators

Neptune Graph Paths (multi-hop):

nexus_find_warm_intro({
  target_company: "...",
  max_hops: 3,
  include_leaddelta: true
})

Returns paths ranked by: hop count (fewer = better) × relationship strength × intro quality.

Phase 4: Combined Scoring

Merge LeadDelta and Neptune paths with unified scoring:

Signal Weight Source
LeadDelta warm intro score 40% LeadDelta AI
Neptune path length (inverse) 25% Graph DB
Relationship recency 20% Both
Mutual connections 15% Both

Intro Quality Classification:

  • Strong (0.75+): Direct connection or 1-hop with high LeadDelta score — recommend immediately
  • Medium (0.50-0.75): 2-hop or moderate LeadDelta score — recommend with context
  • Weak (<0.50): 3+ hops or low scores — surface with caveat

Phase 5: Message Drafting

Draft personalized intro request messages calibrated to the connector's relationship:

Tone calibration (based on nexus_relationship_score):

  • A/A+ grade (85+): Warm, casual — "Hey [name], hope you're well! Quick favor..."
  • B/C grade (40-70): Professional — "Hi [name], I hope this finds you well. I wanted to reach out..."
  • D/F grade (<40): Formal — "Dear [name], I hope you are doing well. I'm writing to..."

Message template:

[Greeting based on relationship grade]

I'm reaching out because I've been trying to connect with [target] 
at [company]. [1 sentence why this matters — deal context or business reason].

Given your [shared context: worked together at X / studied together at Y / 
mutual connection through Z], I thought you might be willing to make an 
introduction.

[Specific ask: Would you be open to a quick intro email? Happy to draft 
the message for you.]

[Closing based on relationship]

Output Format

## Warm Intro Paths: [Target Company/Person]

### Best Path ⭐ (intro quality: strong)
[Connector Name] — [Title] at [Company]
- LeadDelta Score: 0.87/1.0
- Employment overlap: 3.2 years at Goldman Sachs
- Messaging activity: high (last contact 12 days ago)
- Mutual connections: 8

**Suggested Message**:
> "Hey [Connector], hope you're well! I've been trying to connect with [Target] 
> at [Company] re: [context]. Given you both worked at Goldman, would you be 
> open to making an intro? Happy to draft the message."

---

### Path 2 (intro quality: medium)
[Connector Name] — via [Intermediate Person]
- Combined Score: 0.61/1.0
- 2 hops: [You] → [Connector] → [Target]

---

### LeadDelta LinkedIn Network (X additional paths)
[Additional LeadDelta connectors with scores]

---

### No Paths Found?
If 0 paths found: Surface this to user and recommend:
1. Import team LinkedIn connections via `nexus_import_connections`
2. Expand search to 4-5 hops via `nexus_get_intro_paths_between`
3. Consider cold outreach with a highly personalized approach

Error Handling

Scenario Response
LeadDelta API key not configured Surface to user: "LeadDelta connection not configured. Showing Neptune paths only. Configure at /admin/ai-config → Connectors."
LeadDelta API timeout (15s) Fall back to Neptune only, note LeadDelta unavailability
No paths found (0 results both) Recommend LinkedIn import or cold outreach
Target not in any network Suggest broader company search or industry-level search
LeadDelta rate limit hit Queue request, show cached results if available

Integration Points

This skill integrates with:

  • /nexus-network command — primary CLI trigger
  • relationship-graph agent — orchestrates multi-step relationship workflows
  • Nexus CRM — enriched contacts surface in /relationships/contacts
  • Deal flow — warm paths to target companies link to deal records
  • LP Management — relationship paths to LP contacts inform fundraising strategy