linkedin-recon

name: linkedin-recon description: Browser-based LinkedIn profile recon using Browserbase MCP. Visits LinkedIn profiles to extract recent posts, activity, and personalization hooks for outreach. Use when the user types /linkedin-recon or asks to check LinkedIn activity, posts, or content for contacts.

LinkedIn Recon Command

When the user types /linkedin-recon, use Browserbase to visit LinkedIn profiles and extract recent posts, activity, and engagement data for personalized outreach.

Command Pattern

/linkedin-recon <source> where <source> is one of:

If no source is provided, default to accepted (uses the most recent heyreach-accepted-*.csv in clients/partner/<partner>/resources/).

Workflow

Step 1: Identify Profiles to Recon

If accepted (default):

1. Find the most recent heyreach-accepted-*.csv in clients/partner/<partner>/resources/
2. Extract unique LinkedIn URLs from the "LinkedIn URL" column
3. Deduplicate by URL

If single URL:

1. Use the provided URL directly

If CSV path:

1. Read the CSV and extract the "LinkedIn URL" column

Step 2: Create Browserbase Session

1. Call browserbase_session_create to start a browser session
2. This session should ideally be logged into LinkedIn for full profile access
   • If not logged in, public profile data may be limited
   • Inform user if login is needed: "For full profile access, log into LinkedIn in the Browserbase session first."

Step 3: Visit Each Profile

For each LinkedIn profile URL, pace visits to avoid rate limiting:

1. Navigate to activity page: {{linkedin_url}}/recent-activity/all/

   • This shows their posts, shares, and comments
   • Call browserbase_stagehand_navigate with the activity URL

2. Extract activity data: Call browserbase_stagehand_extract with instruction:

   Extract all visible LinkedIn posts and activity. For each post, get:
   - Post content/text (full text)
   - Post date/time
   - Post type (original post, shared post, comment, article)
   - Engagement metrics (likes, comments, shares) if visible
   - Any hashtags used
   - Any companies or people mentioned
   

3. Navigate to profile page: {{linkedin_url}}

   • Call browserbase_stagehand_navigate with the profile URL

4. Extract profile data: Call browserbase_stagehand_extract with instruction:

   Extract the LinkedIn profile information:
   - Full name and headline
   - Current position and company
   - About/summary section
   - Recent experience entries
   - Education
   - Any featured posts or content
   - Skills and endorsements (top ones)
   - Activity summary (number of followers, connections)
   

5. Wait between profiles: Pause 5-10 seconds between profiles to avoid LinkedIn rate limiting

   • Call browserbase_stagehand_navigate to a neutral page or just wait

Step 4: Compile Recon Report

For each contact, generate a recon profile:

## LinkedIn Recon: {{name}} - {{company}}

**Profile**: {{linkedin_url}}
**Headline**: {{headline}}
**Location**: {{location}}

### About
{{about_section_or_"No about section"}}

### Recent Activity (last 30 days)
{{For each post/activity:}}
- **{{date}}** [{{type}}]: {{post_preview_first_100_chars}}
  - Engagement: {{likes}} likes, {{comments}} comments
  - Topics: {{hashtags_or_themes}}

### Personalization Hooks
Based on their activity, here are conversation starters:
1. {{hook_based_on_recent_post_or_activity}}
2. {{hook_based_on_company_news_or_role}}
3. {{hook_based_on_shared_interests_or_events}}

### Suggested Next Message
{{A personalized message draft that references their actual activity/content,
different from the templated sequence they've been ignoring}}

Step 5: Save Report

1. Determine Output Directory:

   • If input was a CSV file: Use {{directory_of_csv}}/recon/
   • If input was accepted or a URL: Use default clients/partner/<partner>/resources/recon/

2. Save individual recon files to the output directory:

   • Filename: {{linkedin-username}}-recon-{{YYYY-MM-DD}}.md
   • Example: zachschofel-recon-2026-02-08.md

3. Save summary CSV with personalization hooks to the parent of output directory:

   • Path: {{output_directory_parent}}/linkedin-recon-{{YYYY-MM-DD}}.csv
   • Columns: First Name, Last Name, Company, Position, LinkedIn URL, Recent Post Topic, Personalization Hook, Suggested Message

4. Display summary:

## LinkedIn Recon Complete

**Profiles scanned**: {{count}}
**With recent activity**: {{active_count}}
**Without activity**: {{inactive_count}}

### Recon Summary
| # | Contact | Company | Posts Found | Top Topic | Hook |
|---|---------|---------|:-----------:|-----------|------|
| 1 | {{name}} | {{company}} | {{count}} | {{topic}} | {{hook_preview}} |

### Individual Reports
- `recon/zachschofel-recon-2026-02-08.md`
- `recon/crystalmoya-recon-2026-02-08.md`
- ...

### Next Steps
- Review hooks and customize messages
- Use `/heyreach-conversations` to see current thread state
- Send personalized follow-ups via HeyReach MCP `send_message`

Step 6: Close Browser Session

1. Call browserbase_session_close to end the session

Rate Limiting & Safety

• Wait 5-10 seconds between profile visits to avoid LinkedIn flagging
• Max 15 profiles per session to stay under the radar
• If more than 15 profiles, split into multiple sessions with a longer break
• If LinkedIn shows a CAPTCHA or block page, stop immediately and inform the user
• Never perform actions on LinkedIn (connect, like, message) - this is READ ONLY recon

Error Handling

• No CSV found: "No accepted connections CSV found. Run /heyreach-export accepted first."
• LinkedIn login required: "Full profile access requires LinkedIn login. Log in via the Browserbase session, then re-run."
• Rate limited/blocked: "LinkedIn is limiting access. Stopping recon. Completed {{count}}/{{total}} profiles. Try again later for the rest."
• Profile not found: "Profile {{url}} returned 404. Skipping."
• No activity found: "No recent posts found for {{name}}. Their profile may be inactive on LinkedIn."

Examples

Recon accepted connections

User: /linkedin-recon
User: /linkedin-recon accepted

→ Finds latest heyreach-accepted CSV
→ Extracts 11 unique LinkedIn URLs
→ Visits each profile's activity page
→ Extracts posts, topics, engagement
→ Generates personalization hooks
→ Saves individual recon files + summary CSV

Recon single profile

User: /linkedin-recon https://www.linkedin.com/in/zachschofel

→ Visits Zach's profile and activity page
→ Extracts posts and profile data
→ Generates personalization hooks
→ Saves recon file

Recon from custom CSV

User: /linkedin-recon clients/partner/<partner>/resources/vip-contacts.csv

→ Reads CSV for LinkedIn URLs
→ Visits each profile
→ Generates recon report