LinkedIn Lead Prospector — Local-Run B2B Lead Generation for LinkedIn

A local-run B2B lead generation workflow for LinkedIn. The buyer runs the skill on their own machine, logs into LinkedIn in their own browser session, and gets a verified, confidence-labeled Excel sheet per run.

What it does

• Searches LinkedIn for prospects by industry, role, location, or company size.
• Cross-verifies named profiles by opening each profile page and confirming name, title, company, and tenure.
• Enriches anonymous profiles ("LinkedIn Member") by matching title + location against the company's team page on their own website.
• Exports a formatted Excel sheet with one row per lead and a confidence column (Verified / Inferred / Low-confidence).

How a typical run looks

1. Read the request — confirm industry / role / geography / list size with the buyer.
2. Buyer starts the Agent — a browser window opens on the buyer's machine.
3. Buyer logs into LinkedIn in their own browser (one-time, session is reused for the run).
4. The skill runs a LinkedIn search, collects profile URLs.
5. For each URL: open profile, extract name, title, company, location, tenure. Tag Verified.
6. For "LinkedIn Member" placeholders: open the company's /team or /about page, fuzzy-match title + location. Tag Inferred (or Low-confidence if no match).
7. Write Excel with one row per lead.

What the buyer needs to provide / install

This is a Download Agent — runs locally on the buyer's machine. Required setup:

• Python 3.11+
• playwright-cli installed and Chromium downloaded
• openpyxl for Excel output
• A working LinkedIn account (the buyer logs in manually the first time)
• ~30-60 minutes per run (cap of ~20-30 profiles per run to stay within LinkedIn rate limits)

What it does NOT do

• Does not send connection requests or InMails on the buyer's behalf.
• Does not scrape at scale (capped ~20-30 profiles per run by design).
• Does not bypass LinkedIn's login wall — the buyer logs in themselves, in their own browser.
• Does not run in Capafy's cloud — by design, this Agent runs locally because cloud browsers cannot reliably authenticate to a buyer's LinkedIn account.

Output format

One .xlsx file per run. Columns:

• name — full name as it appears on the profile (or empty if anonymous)
• title — current job title
• company — current employer
• location — city / region
• linkedin_url — profile URL (or search-result URL if anonymous)
• source_url — page used for verification (profile / team page / company site)
• confidence — Verified / Inferred / Low-confidence
• notes — anything worth flagging (recent job change, multiple roles, etc.)

Important caveats

• For reference only — final list quality depends on the input search criteria, the buyer's existing LinkedIn network visibility, and whether the target company has a public team page. This skill does not guarantee any specific number of verified leads.
• Compliance — the buyer is responsible for complying with LinkedIn's User Agreement and applicable data-protection laws (GDPR, CCPA, Vietnam PDPD, etc.) in their jurisdiction. The skill only automates browsing in the buyer's own logged-in session.
• Rate limits — ~20-30 profiles per run is a soft cap to avoid LinkedIn throttling the buyer's account. Larger lists are split across multiple runs.

Underlying skill

This Agent packages the linkedin-lead-prospecting skill (Hermes / OpenClaw). Full source, decision log, and recovery patterns are in the skill repository.