linkedin-contact-reacher

LinkedIn Contact Reacher Skill

Discover LinkedIn contacts systematically (BFS or DFS), enrich them with email candidates, optionally send personalized connection requests, and save everything locally.

Specialized Agents

Agent	File	When to Use
Contact Discovery Agent	skills/agents/contact-discovery-agent/SKILL.md	BFS/DFS strategy, seed selection, traversal tuning
Outreach Agent	skills/agents/outreach-agent/SKILL.md	Message templates, connection request best practices, rate limits
Email Generator Agent	skills/agents/email-generator-agent/SKILL.md	Email pattern logic, domain inference, validation
QA Agent	skills/agents/qa-agent/SKILL.md	Pre-flight checks, session validation, output quality
Automation Agent	skills/agents/automation-agent/SKILL.md	Timing, retry logic, rate limiting

Script Files

File	Purpose
discoverContacts.js	BFS/DFS traversal — finds contacts from seeds
extractContactInfo.js	Visits profiles, generates email candidates
reachContacts.js	Sends connection requests with personalized notes
saveOutput.js	Saves results to JSON + CSV locally

Prerequisites

• Logged into LinkedIn
• Playwright MCP tools available
• Output directory writable (default: ./output/)

---

Orchestrated Run Order

1. QA Agent          → preFlightCheck(page)
2. Discovery Agent   → discoverContacts()      ← BFS or DFS
3. Extract Agent     → extractContactInfo()    ← enrich + email candidates
4. [Optional] Reach  → reachContacts()         ← send connection requests
5. Save              → saveOutput()            ← JSON + CSV locally
6. QA Agent          → generateReport()

---

Quick Start

Scenario 1: Find referral contacts at a target company

// ── Step 1: Discover contacts via BFS ─────────────────────────────────
// Paste discoverContacts.js, then:

const contacts = await discoverContacts(page, {
  seeds: [
    { type: 'search', company: 'Google', role: 'Engineering Manager' },
    { type: 'search', company: 'Google', role: 'Software Engineer' }
  ],
  strategy: 'bfs',          // breadth-first: broad coverage of the company
  maxContacts: 30,
  maxDepth: 1,              // depth=1: only direct search results, no expansion
  targetCompanies: ['Google'],
  targetRoles: ['engineer', 'manager', 'tech lead'],
  connectionDegree: ['1st', '2nd', '3rd'],
});

// ── Step 2: Enrich with email candidates ──────────────────────────────
// Paste extractContactInfo.js, then:

const enriched = await extractContactInfo(page, contacts, {
  companyDomains: { 'Google': 'google.com' },
  guessPersonalEmail: false,   // set true to also get gmail/outlook guesses
});

// ── Step 3: Save output ───────────────────────────────────────────────
// Paste saveOutput.js, then:

await saveOutput(enriched, {
  format: 'both',
  outputDir: './output',
  label: 'google-referrals'
});
// Output: ./output/google-referrals_2026-04-27T....{json,csv}

// ── Step 4: (Optional) Send connection requests ───────────────────────
// Paste reachContacts.js, then:

const { results, stats } = await reachContacts(page, enriched, {
  purpose: 'referral',
  userProfile: {
    name: 'Your Name',
    role: 'Software Engineer',
    targetCompany: 'Google'
  },
  maxPerSession: 10,      // stay under LinkedIn's limit
});

// Save final results with reach status
await saveOutput(results, { label: 'google-referrals-reached', format: 'both' });

Scenario 2: Networking DFS — deep dive from one person's profile

const contacts = await discoverContacts(page, {
  seeds: [{ type: 'profile', url: 'https://www.linkedin.com/in/someengineer/' }],
  strategy: 'dfs',          // depth-first: explore one thread deeply
  maxContacts: 20,
  maxDepth: 2,
  targetCompanies: ['Stripe', 'Shopify', 'Airbnb'],
  connectionDegree: ['2nd', '3rd'],
});

const enriched = await extractContactInfo(page, contacts, {
  companyDomains: {
    'Stripe':   'stripe.com',
    'Shopify':  'shopify.com',
    'Airbnb':   'airbnb.com',
  },
  guessPersonalEmail: false,
});

await saveOutput(enriched, { label: 'networking-dfs', format: 'both' });

Scenario 3: Multiple companies, networking only (no outreach)

const allContacts = [];

for (const target of [
  { company: 'Stripe', role: 'backend engineer' },
  { company: 'Shopify', role: 'software engineer' },
]) {
  const batch = await discoverContacts(page, {
    seeds: [{ type: 'search', ...target }],
    strategy: 'bfs',
    maxContacts: 15,
    maxDepth: 1,
    targetCompanies: [target.company],
  });
  allContacts.push(...batch);
  await page.waitForTimeout(5000); // pause between companies
}

const enriched = await extractContactInfo(page, allContacts, {
  companyDomains: { 'Stripe': 'stripe.com', 'Shopify': 'shopify.com' }
});

await saveOutput(enriched, { label: 'multi-company', format: 'both' });

---

Configuration Reference

discoverContacts(page, options)

Option	Default	Description
seeds	[]	Array of seed objects: { type: 'search', company, role, keywords } or { type: 'profile', url }
strategy	'bfs'	'bfs' (queue, broad) or 'dfs' (stack, deep)
maxContacts	30	Maximum unique contacts to collect
maxDepth	2	Max traversal depth from seeds. 1 = search results only
targetCompanies	[]	Filter: only keep contacts at these companies (empty = all)
targetRoles	[]	Filter: only keep contacts whose title contains one of these strings
connectionDegree	['1st','2nd','3rd']	Which LinkedIn connection degrees to include
delayMin	3000	Min delay between navigations (ms)
delayMax	6000	Max delay (ms)

extractContactInfo(page, contacts, options)

Option	Default	Description
companyDomains	{}	Map of { 'Company Name': 'domain.com' } for email generation
guessPersonalEmail	false	Also generate gmail/outlook/hotmail variants
delayMin	2500	Min delay between profile visits (ms)
delayMax	5000	Max delay (ms)

reachContacts(page, contacts, options)

Option	Default	Description
purpose	'networking'	'referral' | 'networking' | 'custom'
customMessage	''	Used when purpose='custom' (max 300 chars)
userProfile	{}	{ name, role, targetCompany } — used in message templates
maxPerSession	15	Max connection requests per session (LinkedIn limit: ~100/week)
onlyUnreached	true	Skip contacts already reached
delayMin	4000	Min delay between requests (ms)
delayMax	8000	Max delay (ms)

saveOutput(contacts, options)

Option	Default	Description
format	'both'	'json' | 'csv' | 'both'
outputDir	'./output'	Output directory (Node.js) or download (browser)
label	'contacts'	Filename prefix. Result: label_YYYY-MM-DDTHH-MM-SS.{json,csv}

---

Output Schema

Each contact in the output has:

{
  "name": "Jane Smith",
  "firstName": "Jane",
  "lastName": "Smith",
  "title": "Engineering Manager",
  "company": "Google",
  "location": "San Francisco, CA",
  "linkedinUrl": "https://www.linkedin.com/in/janesmith/",
  "connectionDegree": "2nd",
  "depth": 0,
  "companyDomain": "google.com",
  "emailCandidates": [
    "[email protected]",
    "[email protected]",
    "[email protected]",
    "[email protected]"
  ],
  "personalEmailCandidates": [],
  "publicEmail": null,
  "websites": [],
  "messageSent": false,
  "reachStatus": "",
  "messageText": "",
  "reachedAt": null,
  "via": "seed:search"
}

---

BFS vs DFS — When to Use Which

	BFS	DFS
Goal	Cover many people at a company broadly	Explore one person's extended network deeply
Best for	Referral hunting at a specific company	Networking via a trusted contact's connections
Depth setting	maxDepth: 1 or 2	maxDepth: 2 or 3
Result shape	Wide and shallow — many contacts from search	Narrow and deep — contacts 2–3 hops away
Risk	May include less-relevant matches	May drift far from target company

---

Email Candidate Patterns

Generated email patterns (in priority order):

Pattern	Example
firstname.lastname@domain	[email protected]
flastname@domain	[email protected]
firstnamelastname@domain	[email protected]
firstnamei@domain	[email protected]
firstname@domain	[email protected]
lastname.firstname@domain	[email protected]
firstname_lastname@domain	[email protected]
f.lastname@domain	[email protected]
lastname@domain	[email protected]
firstname-lastname@domain	[email protected]

Verify candidates with tools like Hunter.io or email verification APIs before sending.

---

Safety & Rate Limits

Action	Recommended Limit
Connection requests per day	≤ 20
Connection requests per week	≤ 80 (free) / 150 (Premium)
Profile visits per session	≤ 50
Session duration	< 45 minutes
Break between sessions	2+ hours

LinkedIn may restrict accounts that send too many requests. Start conservatively.

---

Legal & Ethical Usage

• Only contact people you have a genuine professional reason to reach
• Personalize messages — do not spam
• Respect LinkedIn's Terms of Service
• Do not use harvested emails for bulk cold outreach without consent
• Email candidates are guesses only — verify before use