analyze-subject

Files (1)

• SKILL.md

SKILL.md (first lines)

# Subject Analysis Command

Comprehensive individual intelligence gathering combining internal data sources with extensive OSINT to build detailed profiles of professional history, digital footprint, network connections, thought leadership, and behavioral patterns.

## Usage

```
/crm:analyze-subject [Subject Name]
```

**Research Contexts:** Prospect | Interviewer | Investor | Partner | Client | General

---

# EXECUTION WORKFLOW

## Phase 1: Locate or Create Dossier

1. Run `crm_search` for the subject name
2. If found: run `crm_outline` to see structure and fill rates
3. If not found: ask user for category, then use `crm_create` to create the dossier
4. Note the dossier path — you will need it for Phase 6

## Phase 2: Identity Verification

**For common names, STOP and confirm identity before proceeding.**

Confirm these data points match across 2+ sources:
- Full legal name + current title + current organization
- Location consistent with known information
- Email domain matches organization
- LinkedIn URL or unique identifier available

**If identity cannot be confirmed with HIGH confidence, ask user for additional identifying information.**

## Phase 3: Data Collection

Run these searches **in parallel** where possible.

### 3A. Internal Sources

| Source | Method | Extract |
|--------|--------|---------|
| **Gmail** | `mcp__gmail__search_emails` with `"from:[subject] OR to:[subject]"` | Communication patterns, tone, topics, key dates, quotes |
| **Limitless** | `curl -s "https://api.limitless.ai/v1/lifelogs?limit=50&search=[NAME]" -H "X-API-Key: $(printenv LIMITLESS_API_KEY)"` | Meeting transcripts, verbal patterns, in-person dynamics |
| **Local files** | Grep for subject name in `04_RELATIONSHIPS/CRM/`, `03_PROFESSIONAL/Meetings/`, `01_FINANCIAL/` | Existing analysis, meeting notes, transcripts |
| **CRM** | `crm_read` on existing dossier sections | Prior intelligence, interaction history |

**Transcript reading rule:** Read entire files. For files >2000 lines, use the Read tool with offset/limit to chunk through sequentially. Never sample or skip sections.

### 3B. OSINT — Digital Footprint

Search each platform systematically. Run searches in parallel batches.

**Platform checklist:**
- [ ] LinkedIn — profile, connections, activity, career gaps, endorsements
- [ ] Twitter/X — handle, follower count, content focus, notable posts, red flags
- [ ] Personal website — bio, portfolio, services, blog
- [ ] Substack/Newsletter — publication, frequency, themes, audience
- [ ] Medium/Blog — articles, publications, content themes
- [ ] YouTube — channel, appearances, interviews, conference talks
- [ ] GitHub — repos, activity, technical focus (if applicable)
- [ ] Instagram/Facebook/Threads — public presence only

**Search pattern for each platform:**
```
WebSearch: "[Subject Name]" site:[platform.com]
WebSearch: "[Subject Name]" "[Company]" [platform]
WebFetch: [profile URL if found]
```

### 3C. OSINT — Thought Leadership

- [ ] Books, academic papers, whitepapers (Google Scholar, Amazon)
- [ ] Speaking engagements (keynote, panel, workshop)
- [ ] Podcast appearances (Apple Podcasts, Spotify)
- [ ] Media coverage (Bloomberg, WSJ, TechCrunch, industry press)
- [ ] Op-eds, industry articles

### 3D. OSINT — Professional Network

- [ ] Board and advisory roles (Crunchbase, company websites)
- [ ] Investment activity (AngelList, Crunchbase, portfolio companies)
- [ ] Professional associations and memberships
- [ ] Co-founder and partnership history
- [ ] Key collaborators and frequent connections

### 3E. OSINT — Due Diligence

- [ ] Court records and litigation (`"[Name]" lawsuit OR litigation OR "v."`)
- [ ] Regulatory actions (`"[Name]" SEC enforcement OR FINRA OR disciplinary`)
- [ ] Corporate registry filings (LLC, Inc, Corp)
- [ ] SEC filings (for public company executives)
- [ ] Patent and trademark filings
- [ ] Credential verification (degrees, licenses, certifications — search official registries)

## Phase 4: Verification

Cross-reference key claims across multiple sources. For each major factual claim:

1. **Classify confidence:**
   - VERIFIED — confirmed by 2+ authoritative sources
   - CORROBORATED — supported by 1 additional source
   - UNVERIFIED — single source only
   - CONFLICTING — contradicting information found
   - REFUTED — evidence contradicts the claim

2. **Escalation rules:**
   - REFUTED credential claims → increase risk score by 2
   - 3+ CONFLICTING claims → flag for manual review
   - UNVERIFIED claims → include with `[UNVERIFIED]` tag

3. **High-priority verification** (always attempt):
   - Professional licenses → licensing body registry
   - Educational degrees → university alumni records
   - Employment claims → company announcements, SEC filings
   - Legal outcomes → court records

## Phase 5: Analysis

From reggiechan74/crm-mcp-server, MIT.

More from reggiechan74/crm-mcp-server