Check Mutual Connections

Store this — it's used in the final output header to show whose perspective these mutual connections are from.

Step 3: Navigate to the Profile

Open the target's LinkedIn profile in the browser. The user must already be logged in.

If the page shows a login wall, stop and tell the user.

Step 4: Open Mutual Connections

After the profile loads, find the mutual connections link — it reads something like "Sammy Abdullah, Mark Tornetta, and 11 other mutual connections". Click it.

This opens a search results page filtered to the target with 1st-degree connections checked by default.

Step 5: Enable 2nd-Degree Filter FIRST

Important: Enable 2nd degree before extracting any results. This lets you collect all connections (1st and 2nd) in a single pass instead of paginating twice.

On the search results page, find the 2nd-degree connection filter (a radio button or checkbox labeled "2nd"). Click it so both 1st and 2nd are checked simultaneously.

Wait 2 seconds for results to reload.

Why this order matters: The default search shows only 1st degree connections. If you extract 1st degree first and then add the 2nd degree filter, the results reshuffle completely — 1st degree connections reappear on later pages, causing duplicate work. Enabling both filters upfront avoids this.

Pagination warning: LinkedIn dynamically expands page count as you go deeper. A search that initially shows 10 pages may grow to 15, 17, 18+ as you paginate. This is normal — keep going until "Next" disappears.

Step 6: Extract Connections

For each page of results, extract:

• Name
• LinkedIn profile URL (the /in/ link)
• Connection degree (1st or 2nd — parse from • 1st or • 2nd in the link text)
• Title (job title / headline)

Preferred: JavaScript evaluation in the browser context:

JSON.stringify(
  Array.from(document.querySelectorAll('a[href*="/in/"]'))
    .map(a => {
      var text = a.textContent.trim().substring(0, 400);
      var url = a.href.split('?')[0];
      var degree = text.includes('2nd') ? '2nd' : text.includes('1st') ? '1st' : null;
      return { text: text, url: url, degree: degree };
    })
    .filter(a => a.url.includes('/in/') && !a.url.includes('/in/{target_slug}') && a.degree)
    .reduce((acc, a) => { if (!acc.some(x => x.url === a.url)) acc.push(a); return acc; }, [])
)

This captures degree info from the link text in a single pass, so you don't need separate extraction logic per degree.

Fallback: Parse the accessibility tree / snapshot. Each result entry contains name, degree, title, location. Profile URLs can be extracted by reading link hrefs.

Step 7: Paginate

Check for "Next" or page number buttons. If they exist, click "Next", wait 2 seconds, repeat step 6.

Continue until "Next" is absent. Note: LinkedIn may show up to 10 page buttons at a time — new ones appear as you advance, so don't assume the highest visible page number is the last.

Step 8: Deduplicate, Format, and Output

Combine all pages. Deduplicate by LinkedIn URL. Sort: 1st degree first, then 2nd, alphabetical within each group.

Output Format

*Mutual Connections: {Logged-In User Name} ↔ {Target Name}*

*{Target Name}*
{Title} | {Location}
{linkedin_url}

*1st Degree*
- <{url}|{Name}> — {Title}
- <{url}|{Name}> — {Title}

*2nd Degree*
- <{url}|{Name}> — {Title} (via {mutual name} + {N} other mutuals)
- <{url}|{Name}> — {Title} (via {mutual name} + {N} other mutuals)

Rules:

• First line is a header showing whose mutual connections these are: the logged-in user's name and the target's name, connected with ↔
• Next block is the target's name in bold, followed by their title, location, and LinkedIn URL
• Group connections under *1st Degree* and *2nd Degree* headers
• Each connection is: - <linkedin_url|Name> — Title
• For 2nd degree connections, append (via {mutual name} + {N} other mutuals) if mutual connection info is available from the search results
• Do NOT include a status line (e.g. "Connect button available") — just the connections
• Include LinkedIn URLs for every connection that has one
• Keep messages under ~4000 chars; split into multiple if needed

Error Handling

• Login wall — stop, tell user to log in
• No mutual connections link — target may have restricted visibility, inform user
• Empty results — report "No mutual connections found"
• Rate limit / CAPTCHA — stop immediately, alert user