LinkedIn Skills for AI Agents: Claude Code, Codex, Cursor, and Windsurf

Source: github.com/Linked-API/linkedin-skills

Powered by: Linked API (linkedapi.io) + linkedin-cli (@linkedapi/linkedin-cli)

License: MIT

Works with: Claude Code, Codex, Cursor, Windsurf, OpenClaw, and any AI agent that can execute shell commands

LinkedIn automation skills for AI agents. One skill folder. Covers profiles, people search, company research, messaging, connections, posts, reactions, comments, Sales Navigator, and analytics.

How it works: the linkedin CLI sends requests to Linked API, which runs a dedicated cloud browser with a residential IP for each connected LinkedIn account. No local browsers. No proxies. No Selenium. No infrastructure to manage.

What this is and how it differs

Most LinkedIn automation tools fall into two categories:

1. Browser extensions that require you to have Chrome open and logged in
2. Unofficial API wrappers that break whenever LinkedIn updates its scraping defences

Linked API runs a real cloud browser instance for each connected account. When your agent runs a command, the CLI sends the request to Linked API's infrastructure, where an isolated browser performs the action on LinkedIn, simulating human-like mouse movements, keyboard input, and natural browsing patterns.

The result: your AI agent can run LinkedIn actions without a browser open on your machine, without API credentials that expire, and without scraping code that breaks every few weeks.

Operations are not instant. Expect 30 seconds to several minutes depending on complexity. This is intentional: cloud browser operations take time because they simulate human behaviour.

Install

# Step 1: Install the CLI
npm install -g @linkedapi/linkedin-cli

# Step 2: Set up your account (get tokens at app.linkedapi.io)
linkedin setup

# Step 3: Copy the skill into your agent
# Project-level (this project only):
cp -R linkedin-skills/linkedin .claude/skills/linkedin

# Global (all projects):
cp -R linkedin-skills/linkedin ~/.claude/skills/linkedin

Authentication

1. Go to app.linkedapi.io and sign up or log in
2. Connect your LinkedIn account
3. Copy the Linked API Token and Identification Token from the dashboard
4. Run: linkedin setup --linked-api-token=TOKEN --identification-token=TOKEN

If a command fails with exit code 2 (authentication error), the tokens need to be refreshed. Run linkedin setup again.

Exit code 0 means the API call succeeded. Always check the success field in the JSON response for the action outcome. Non-zero exit codes indicate infrastructure errors.

Command categories

Category	Commands	What it covers
People	person fetch, person search	Profile data, experience, education, skills, posts, reactions, comments
Companies	company fetch, company search	Company profile, employees, decision makers, posts
Messaging	message send, message list, message read	DMs and InMail
Connections	connection send, connection list, connection remove	Managing your 1st-degree network
Posts	post create, post react, post comment	Publishing and engaging with content
Sales Navigator	navigator person/company fetch and search	Sales Nav profiles, companies, DMs
Stats	stats ssi, stats performance, stats usage	Social Selling Index, analytics, API usage

Person commands. Fetch full profiles, search for people by role, company, school, industry, and location. All outputs in structured JSON.

linkedin person fetch

What it does

Copy-paste prompts

GTM use case: account research before outreach

linkedin person search

What it does

Copy-paste prompts

GTM use case: ICP list building

Company commands. Fetch company profiles, find employees and decision makers, pull recent posts for buying signals.

linkedin company fetch

What it does

Copy-paste prompts

GTM use case: account brief before outreach

linkedin company search

What it does

Copy-paste prompts

Messaging and connection commands. Send DMs, read conversations, manage your connection network. The most powerful GTM use case: research then message in one agent session.

linkedin message send

What it does

Copy-paste prompt

linkedin connection send

What it does

Copy-paste prompt

linkedin post create

What it does

Copy-paste prompt

linkedin post comment and post react

What it does

Copy-paste prompt (warmup sequence)

Sales Navigator commands. Requires a LinkedIn Sales Navigator subscription. Uses hashed URLs for person and company lookups. More powerful filters than standard LinkedIn search.

linkedin navigator person search

What it does

Copy-paste prompts

linkedin navigator company search

What it does

Copy-paste prompts

linkedin navigator company fetch

What it does

Copy-paste prompts

linkedin navigator message send

What it does

Copy-paste prompt

Stats commands and chained GTM workflows combining multiple commands in one agent session.

linkedin stats

Copy-paste prompts

Chained workflow 1: ICP list to connection requests

Build an ICP prospect list and send connection requests.

Step 1: Search for ICP-matching profiles
linkedin person search \
  --term "[TARGET TITLE]" \
  --locations "[GEOGRAPHY]" \
  --limit 20 \
  --json -q

Step 2: For each result, fetch the profile to confirm ICP fit
linkedin person fetch [URL] --experience --posts --posts-limit 2 --json -q

Step 3: Score each against ICP: [CRITERIA]. Keep only Tier 1 and Tier 2.

Step 4: For each kept profile, write a connection note under 300 characters referencing:
- Their most recent post topic, OR
- Their current role and a specific pain you solve

Step 5: Show me the full list and notes. Wait for my approval.

Step 6: For each approved:
linkedin connection send [URL] --note "[NOTE]" --json -q

Log to CSV: name, company, title, URL, note, date sent.

Chained workflow 2: Company signal to DM

Find companies showing a buying signal and reach the decision maker.

Step 1: Search for companies
linkedin company search --term "[INDUSTRY]" --sizes "[SIZE RANGE]" --json -q

Step 2: For each company, fetch the profile with decision makers and posts
linkedin company fetch [URL] --dms --posts --json -q

Step 3: Check for buying signals in their recent posts:
- Hiring signals: job postings for roles that indicate they need what you sell
- Content signals: posts about the problem you solve
- Growth signals: expansion announcements, funding, new markets

Step 4: For companies with signals, get the decision maker profile
linkedin person fetch [DM URL] --experience --posts --posts-limit 3 --json -q

Step 5: Write a connection note referencing the specific signal

Step 6: linkedin connection send [URL] --note "[NOTE]" --json -q

Chained workflow 3: Daily LinkedIn session

Run my daily LinkedIn outreach session.

Step 1: Check performance
linkedin stats performance --json -q
Return: profile views vs last week, post impressions, search appearances

Step 2: Check SSI score
linkedin stats ssi --json -q
Flag if any dimension dropped more than 5 points

Step 3: Find [N] new ICP prospects
linkedin person search --term "[TITLE]" --locations "[LOCATION]" --limit [N] --json -q

Step 4: For the top 5 ICP matches, build account briefs
linkedin person fetch [URL] --posts --posts-limit 3 --experience --json -q (x5)

Step 5: Write connection notes for each. Show me all 5. Wait for approval.

Step 6: Send approved requests
linkedin connection send [URL] --note "[NOTE]" --json -q

Step 7: Log all actions to daily-outreach-log.csv
Report: N searches run, N profiles fetched, N connection requests sent.

AI Driven, Personalized Outreach on LinkedIn: Prosp 💜

↳ Start a 14 Day Free Trial 💜 ↳ Watch Full Tutorial ▶️