Universal Deep Search

Comprehensive intelligence gathering across ALL available data sources with cascading analysis.

Core Principle: CASCADING SEARCH

Every search expands to related entities:

• PERSON → + their company + key colleagues + company news
• COMPANY → + founders + C-level team + investors + competitors mentions
• TOPIC → + key people mentioned + companies involved + YC startups in space

Query Classification

Type 1: PERSON

Triggers: Names, roles ("CEO of"), profile URLs, "who is", "find person"

Context to extract:

• Full name (required)
• Company (helpful)
• Role/title (helpful)
• Location (helpful)

Type 2: COMPANY

Triggers: Company names, domains (.io, .com), "startup", "company", linkedin.com/company/

Context to extract:

• Company name (required)
• Domain (helpful)
• Industry (helpful)
• Location (helpful)

Type 3: TOPIC

Triggers: Abstract concepts, questions, hashtags, "news about", "trends in"

Context to extract:

• Keywords (required)
• Time frame (helpful)
• Industry/niche (helpful)

PERSON Search Workflow

Phase 1: Find & Identify Person

1. search_linkedin_users
   - keywords: "[name]"
   - company_keywords: "[company]" if known
   - title: "[role]" if known
   - count: 10

2. If multiple matches → present top 5 for confirmation
3. If YC founder suspected → search_yc_founders(query="[name]")

Phase 2: Deep Profile Analysis

1. get_linkedin_profile
   - user: "[username or URL]"
   - with_experience: true
   - with_education: true
   - with_skills: true
   
   CRITICAL: Extract and save:
   - Full URN: urn:li:fsd_profile:ACoAAA...
   - Current company slug/URN
   - Current role & start date

2. get_linkedin_user_posts
   - urn: "[full URN from above]"
   - count: 50
   
3. get_linkedin_user_comments
   - urn: "[full URN]"
   - count: 30

4. get_linkedin_user_reactions
   - urn: "[full URN]"
   - count: 50

Phase 3: Cross-Platform Presence

1. search_twitter_users
   - query: "[name] [company]"
   - count: 5
   
2. get_twitter_user (if found)
   - user: "[handle]"

3. get_twitter_user_posts
   - user: "[handle]"
   - count: 100

4. search_reddit_posts
   - query: "[name] OR [username]"
   - count: 20

5. get_instagram_user (if B2C/personal brand)
   - username: "[handle]"

6. duckduckgo_search
   - query: "[name] [company] speaker OR interview OR article OR podcast"
   - count: 10

7. duckduckgo_search
   - query: "[name] site:github.com OR site:medium.com OR site:substack.com"
   - count: 10

Phase 4: CASCADE → Company Analysis

Always analyze person's current company:

1. get_linkedin_company
   - company: "[company slug from profile]"

2. get_linkedin_company_posts
   - urn: "[company URN]"
   - count: 20

3. parse_webpage
   - url: "[company website]"
   - only_main_content: true

4. parse_webpage
   - url: "[company website]/about"

5. search_yc_companies (check if YC company)
   - query: "[company name]"

6. duckduckgo_search
   - query: "[company] funding news 2024 2025"
   - count: 10

Phase 5: CASCADE → Key Colleagues

1. get_linkedin_company_employees
   - companies: ["[company slug]"]
   - keywords: "founder OR CEO OR CTO OR VP"
   - count: 10

2. For top 2-3 executives:
   - get_linkedin_profile (brief)

COMPANY Search Workflow

Phase 1: Find & Identify Company

1. search_linkedin_companies
   - keywords: "[company name]"
   - location: "[location]" if known
   - industry: "[industry]" if known
   - count: 10

2. search_yc_companies
   - query: "[company name]"
   - hits_per_page: 20

3. If multiple matches → present options for confirmation

Phase 2: Deep Company Profile

1. get_linkedin_company
   - company: "[slug]"
   
   Extract: URN, employee count, industry, description

2. get_linkedin_company_employee_stats
   - urn: "[company URN]"

3. get_linkedin_company_posts
   - urn: "[company URN]"
   - count: 30

Phase 3: Website Intelligence

1. parse_webpage (homepage)
   - url: "https://[domain]"
   - only_main_content: true
   - extract_contacts: true

2. parse_webpage (about)
   - url: "https://[domain]/about"

3. parse_webpage (pricing)
   - url: "https://[domain]/pricing"

4. parse_webpage (team/leadership)
   - url: "https://[domain]/team" OR "/about#team"

5. get_sitemap
   - url: "https://[domain]/sitemap.xml"
   - count: 50

Phase 4: Social & Community Presence

1. search_twitter_users
   - query: "[company name]"
   - count: 5

2. get_twitter_user
   - user: "[company handle]"

3. get_twitter_user_posts
   - user: "[handle]"
   - count: 50

4. search_twitter_posts
   - query: "[company] OR @[handle]"
   - count: 100

5. search_reddit_posts
   - query: "[company name]"
   - count: 50

6. search_reddit_posts
   - query: "[company name]"
   - subreddit: "[relevant sub]" (e.g., "startups", "SaaS", industry-specific)
   - count: 30

7. search_instagram_posts (if B2C)
   - query: "#[company] OR [company name]"
   - count: 20

Phase 5: CASCADE → Leadership Team Analysis

Always analyze founders and C-level:

1. get_linkedin_company_employees
   - companies: ["[slug]"]
   - keywords: "founder"
   - count: 10

2. get_linkedin_company_employees
   - companies: ["[slug]"]
   - keywords: "CEO OR CTO OR CPO OR CFO OR COO"
   - count: 10

3. For each founder/C-level (top 5):
   a. get_linkedin_profile
      - with_experience: true
      - with_education: true
      - with_skills: true
   
   b. get_linkedin_user_posts
      - count: 20
   
   c. search_twitter_users → get_twitter_user_posts
      - count: 30

4. search_yc_founders
   - query: "[founder names]"

Phase 6: News & External Intelligence

1. duckduckgo_search
   - query: "[company] funding news"
   - count: 10

2. duckduckgo_search
   - query: "[company] launch product announcement"
   - count: 10

3. duckduckgo_search
   - query: "[company] review OR competitor OR alternative"
   - count: 10

4. parse_webpage (top news articles)
   - Parse 3-5 most relevant results

5. If tech company:
   duckduckgo_search
   - query: "[company] site:github.com"
   - count: 5

Phase 7: Y Combinator Check

1. search_yc_companies
   - query: "[company name]"

2. If found:
   get_yc_company
   - company: "[slug]"
   
   Extract: batch, status, funding, team size, founders

3. search_yc_founders
   - query: "[company name]"

TOPIC Search Workflow

Phase 1: Web Overview

1. duckduckgo_search
   - query: "[topic keywords]"
   - count: 15

2. duckduckgo_search
   - query: "[topic] trends 2024 2025"
   - count: 10

3. duckduckgo_search
   - query: "[topic] news recent"
   - count: 10

4. parse_webpage
   - Parse top 5 most authoritative results

Phase 2: Professional Discussion (LinkedIn)

1. search_linkedin_posts
   - keywords: "[topic]"
   - count: 30

2. search_linkedin_companies
   - keywords: "[topic] OR [related terms]"
   - count: 20

3. search_linkedin_users
   - keywords: "[topic] expert OR thought leader"
   - count: 10

Phase 3: Real-time Sentiment (Twitter)

1. search_twitter_posts
   - query: "[topic]"
   - count: 100

2. search_twitter_posts
   - query: "[topic] #[hashtag]"
   - count: 50

3. search_twitter_users
   - query: "[topic] expert"
   - count: 10

Phase 4: Community Insights (Reddit)

1. search_reddit_posts
   - query: "[topic]"
   - count: 50

2. search_reddit_posts
   - query: "[topic]"
   - subreddit: "[most relevant sub]"
   - count: 30

3. get_reddit_post_comments (on popular posts)
   - Parse top 3 most discussed threads

Phase 5: Visual Content (Instagram)

1. search_instagram_posts
   - query: "#[topic_hashtag]"
   - count: 20

Phase 6: Startup Landscape (Y Combinator)

1. search_yc_companies
   - query: "[topic]"
   - industries: ["[related industry]"]
   - hits_per_page: 50

2. For top 5 relevant YC companies:
   get_yc_company
   - company: "[slug]"

3. search_yc_founders
   - query: "[topic]"
   - industries: ["[related industry]"]

Phase 7: CASCADE → Key Entities

Identify and analyze key people/companies mentioned:

1. From all collected data, extract:
   - Most mentioned people → run PERSON mini-analysis
   - Most mentioned companies → run COMPANY mini-analysis

2. Mini-analysis (for each top entity):
   - LinkedIn profile/company
   - Recent posts
   - Twitter presence

Validation & Cross-Referencing

For PERSON

• Name matches across platforms
• Company/role consistency
• Profile photo verification (same person)
• Timeline consistency (career progression)

For COMPANY

• Domain matches LinkedIn company
• Employee count consistency
• Founding date alignment
• Industry classification match

For TOPIC

• Source authority ranking
• Recency weighting
• Cross-source fact verification
• Sentiment consistency

Confidence Scoring

• HIGH: 4+ validation points, consistent across 3+ platforms
• MEDIUM: 2-3 validation points, minor inconsistencies
• LOW: 1 validation point or significant conflicts

Output Format

# Deep Search Report: [Subject]

**Type:** PERSON / COMPANY / TOPIC
**Query:** [Original query]
**Confidence:** HIGH / MEDIUM / LOW
**Platforms Searched:** [list all]
**Total API Calls:** [number]
**Analysis Date:** [date]

---

## Executive Summary

[3-5 key findings in bullet points]

---

## Primary Subject Analysis

### [Subject Name]

[Detailed findings organized by data type]

**Profile:**
[Core facts]

**Activity Analysis:**
[Posts, engagement patterns]

**Cross-Platform Presence:**
[Twitter, Reddit, Instagram, Web findings]

---

## Cascaded Analysis

### [Related Entity 1: Company/Person]
[Key findings from cascade]

### [Related Entity 2: Leadership/Colleagues]
[Key findings from cascade]

---

## Topic/Industry Context

[For PERSON/COMPANY: industry context]
[For TOPIC: full analysis here]

### YC Landscape
[Relevant YC companies and founders]

### Recent News & Trends
[From web search]

---

## Cross-Platform Synthesis

**Consistent Facts:**
- [Facts verified across multiple sources]

**Platform-Specific Insights:**
- LinkedIn: [professional persona]
- Twitter: [public opinions]
- Reddit: [community engagement]
- Instagram: [personal brand]

**Inconsistencies/Flags:**
- [Any conflicts to note]

---

## Sources

| # | Platform | URL | Type | Freshness |
|---|----------|-----|------|-----------|
| 1 | LinkedIn | [link] | Profile | Current |
| 2 | Twitter | [link] | Posts | Last 30d |
| 3 | Reddit | [link] | Discussion | Last 7d |
| 4 | YC | [link] | Company | Current |
| 5 | Web | [link] | Article | [date] |

---

## Metadata

- Platforms: LinkedIn, Twitter, Reddit, Instagram, YC, Web
- Data Points Collected: ~[number]
- Validation Status: PASSED / PARTIAL / NEEDS_REVIEW
- Cascade Depth: [how many related entities analyzed]

Quick Reference: All Endpoints Used

LinkedIn (24 tools)

• searchlinkedinusers - find people
• searchlinkedincompanies - find companies
• searchlinkedinsalesnavigatorusers - advanced people search
• getlinkedinprofile - person details
• getlinkedincompany - company details
• getlinkedincompany_employees - team members
• getlinkedincompany_posts - company content
• getlinkedincompanyemployeestats - growth data
• getlinkedinuser_posts - person's posts
• getlinkedinuser_comments - person's comments
• getlinkedinuser_reactions - person's reactions
• searchlinkedinposts - topic search

Twitter (5 tools)

• searchtwitterusers - find accounts
• gettwitteruser - profile details
• gettwitteruser_posts - tweets
• searchtwitterposts - topic/mention search

Instagram (8 tools)

• getinstagramuser - profile
• getinstagramuser_posts - posts
• searchinstagramposts - hashtag/topic search
• getinstagramuser_friendships - followers/following

Reddit (3 tools)

• searchredditposts - find discussions
• getredditpost - post details
• getredditpost_comments - comments

Y Combinator (3 tools)

• searchyccompanies - find startups
• getyccompany - company details
• searchycfounders - find founders

Web (3 tools)

• duckduckgo_search - web search
• parse_webpage - extract content
• get_sitemap - discover pages

Error Handling

No results: Try alternative spellings, broader terms, different platforms Rate limits: Continue with available data, note gaps Private profiles: Note as limitation, use public data Multiple matches: Present options, ask for confirmation

Search Depth Options

User can request:

• Quick scan (10 min): Primary endpoints only, no cascade
• Standard (20-30 min): Full workflow, 1-level cascade (DEFAULT)
• Deep dive (45-60 min): Extended counts, 2-level cascade, all platforms