linkedin-prospector

LinkedIn Prospector (The Hunter)

Identifies high-value prospects by scanning LinkedIn pages and filtering them against your defined Ideal Customer Profile (ICP). Designed to work with Claude for Chrome or an Agentic Browser.

Configuration (Dynamic ICP)

• Reads: data/2-Domaines/sales-profile.json
• Keys:
  • target_geography: List of allowed countries (e.g., ["France", "Canada"]). STRICT FILTER.
  • icp.role_keywords: Titles to match (e.g., ["CTO", "VP Sales"]).
  • icp.company_size: Size range (e.g., "11-50", "51-200").
  • icp.pain_points: Keywords for finding problem-aware prospects.

Triggers

• "find prospects"
• "scan feed"
• "check inbound"
• "who viewed my profile"

Workflow Strategies

1. Outbound Screening (Search & Comments)

Where to run: LinkedIn Search Results, Comment Sections of Thought Leaders. Logic:

1. Scrape: Read the DOM for Name, Headline, Location.
2. Geo Filter: IF Location NOT IN target_geography → SKIP.
3. Role Filter: IF Headline does NOT contain icp.role_keywords → SKIP.
4. Pain Signal: IF text matches icp.pain_points → HIGH PRIORITY.

2. Inbound Screening (Lurker Detection)

Where to run: /me/profile-views/, /mynetwork/invitation-manager/, Post Reactions. Logic:

• Why: Inbound signals (views/likes) indicate intent.
• Fast-Track: If they viewed you + match ICP, they are "Warm" (1-2 touches).

3. Competitor Monitoring

Where to run: Competitor's Post Comments. Logic: Users complaining on competitor posts are high-intent buyers.

Browser Instructions (DOM Scraping)

When reading a page, look for these specific elements:

A. Search Results / Feed:

• Container: .entity-result__item or .feed-shared-update-v2
• Name: .entity-result__title-text or .update-components-actor__name
• Headline: .entity-result__primary-subtitle or .update-components-actor__description
• Location: .entity-result__secondary-subtitle
• Action: Scroll down to load more items before scraping.

B. Profile Page:

• Name: .pv-text-details__left-panel h1
• Headline: .pv-text-details__left-panel .text-body-medium
• Location: .pv-text-details__left-panel .text-body-small
• About: #about section text.
• Experience: #experience section (Current Role).

C. Comments Section:

• Container: .comments-comment-item
• Author: .comments-post-meta__name-text
• Text: .comments-comment-item__main-content
• Headline: .comments-post-meta__headline

Execution Steps

1. Context: Agent checks current URL to determine mode (Search vs Profile vs Inbound).
2. Scrape: Extract list of candidates from DOM.
3. Filter (In-Memory): Apply sales-profile.json rules.
4. Dedup: Check data/2-Domaines/LinkedIn/prospects.md (by Profile URL).
5. Save: Append new matches to prospects.md.

Output Format (prospects.md)

Append to data/2-Domaines/LinkedIn/prospects.md:

Name	Role	Company	Location	Source	Status	Notes
[Name]	[Role]	[Company]	[City, Country]	[Source URL/Comment]	Lead	High Fit; Pain: "CRM is slow"

Example Interaction

User: "Find prospects in these comments" Agent:

1. Reads sales-profile.json → Target: "France", Role: "CTO".
2. Scans 15 comments.
3. Identifies "Jean Dupont, CTO @ TechFlow, Paris". -> MATCH.
4. Identifies "Alice Smith, Recruiter, London". -> DISCARD (Geo/Role mismatch).
5. Saves Jean Dupont.
6. Output: "Found 1 prospect: Jean Dupont (CTO). Saved to prospects.md."

Data Privacy & Safety

• Rate Limit: Max 50 profile views/day (simulated).
• Delay: Wait 2-5 seconds between navigations if automating clicks.
• Privacy: Do not scrape email unless visible on public profile.