linkedin-engine

Run your entire LinkedIn motion inside Claude Code. Build authority with content. Warm every target before you message them. Turn warm conversations into booked calls. One repo.

2023: 0 replies from my DMs. 2026: my whole LinkedIn runs on one repo.

29 skills across three jobs, built from 7 source playbooks, wired to a 13-tool MCP layer and a 17-command CLI. The skills do the research, the writing, and the triage at machine speed. You keep your hand on every message that goes out.

/linkedin-post-writer        Draft a viral-ready post with 10 hook formulas
/linkedin-engager-analytics  See who engaged, ranked by ICP match
/linkedin-outreach           Find the buyer, draft the note, send (20/day cap)
/linkedin-sell-by-chat       Run the 6-message framework from warm to booked

Draft-first by design. Every skill drafts. You approve. Nothing posts or sends on its own. The optional send layer uses your own LinkedIn account and is hard-capped at 20 actions a day to keep new accounts safe. No scraping. No bots. No magic growth.

The 60-second start

# 1. Add the engine to Claude Code
/plugin marketplace add styfinity/linkedin-engine
/plugin install linkedin-engine@styfinity

# 2. Build your brief once (ICP, voice, offer, buyer pains)
/linkedin-onboard

# 3. Write your first post
/linkedin-post-writer

That is the whole setup. A SessionStart hook loads your brief into every session from then on, so every skill writes in your voice, for your ICP, with your offer, without you re-explaining anything.

Want to try before you install? claude --plugin-dir ./linkedin-engine

Related MCP server: @maheidem/linkedin-mcp

How the three parts chain

This is not 29 disconnected prompts. It is one loop that compounds: the content that earns attention feeds the warming list, the warming earns recognition, and the recognition is what makes the outreach land.

flowchart LR
    C["📝 CONTENT<br/>build authority"] --> A["👀 Audience<br/>engages your post"]
    A --> W["🔥 WARMING<br/>score + warm the ICP"]
    W --> O["✉️ OUTREACH<br/>open on a real signal"]
    O --> B(["📅 Booked calls"])
    A -. "warm list" .-> W
    B -. "proof + referrals" .-> C

    style C fill:#0E1B3C,stroke:#06B6D4,color:#fff
    style W fill:#0E1B3C,stroke:#06B6D4,color:#fff
    style O fill:#0E1B3C,stroke:#06B6D4,color:#fff
    style B fill:#06B6D4,stroke:#06B6D4,color:#06223a
    style A fill:#13224A,stroke:#7C3AED,color:#fff

The 5 you will live in

All 29 skills

Plus a SessionStart hook that loads your brief into every session automatically.

What is under the hood

flowchart TD
    Brief["🧠 linkedin-brief.md<br/>ICP · voice · offer · pains"] --> Skills
    subgraph Skills["29 Skills running in Claude Code"]
      direction LR
      Content["📝 10 Content"]
      Warming["🔥 8 Warming"]
      Outreach["✉️ 9 Outreach"]
    end
    Skills --> Approve{"✅ You approve<br/>every draft"}
    Approve --> CLI["⌨️ 17-command CLI"]
    Approve --> MCP["🔌 13-tool MCP layer"]
    CLI --> LI[("🔵 Your LinkedIn account<br/>hard cap: 20 actions / day")]
    MCP --> LI

    style Brief fill:#13224A,stroke:#7C3AED,color:#fff
    style Skills fill:#0A1024,stroke:#26396B,color:#9FB1D2
    style Approve fill:#0E1B3C,stroke:#10B981,color:#fff
    style CLI fill:#0E1B3C,stroke:#06B6D4,color:#fff
    style MCP fill:#0E1B3C,stroke:#06B6D4,color:#fff
    style LI fill:#06B6D4,stroke:#06B6D4,color:#06223a

The skills are the brain. They research, score, draft, and triage. They work with zero setup off pasted data, and they never touch your account on their own.

The CLI is the hands. 17 commands, dependency-free (Node 18+), dry-run by default. It pulls live data when you connect your own account, queues what the skills drafted, and sends under the daily cap.

li doctor               check connection, brief, and today's send budget
li engagers <post>      who engaged a post        li find --title .. --seniority ..
li queue connection ..  queue what the skills drafted
li send-connections --send    send, hard-capped at the daily limit
li tracker · li budget · li triage · li inbox · li log   ... and more (li help)

The MCP layer is the live wire. 13 tools, opt-in. Connect a LinkedIn MCP server with your own account and the skills go live: pull who engaged a post, a target's signals, your inbox, and send the actions you approve. Two of the 13 tools send anything, and both run through the same daily cap.

The 7 source playbooks

The skills are the automation. The playbooks are the thinking they run on. Read them even if you never run a skill.

1. The Governing Principle - sell the operator, not the artefact

2. Post Anatomy and the 10 Hook Formulas

3. The Humanizer Rules - writing that does not read like AI

4. The Warming Doctrine - recognition before the message

5. The Harvest Playbook - viral post to booked call

6. Sell By Chat - the 6-message framework and setter discipline

7. The Conversion Architecture - the layered withhold

The honest part

There is no compliant way to blast thousands of strangers, and LinkedIn bans accounts that move too fast. So this is built warm-first. The skills make the research, the writing, and the triage fast. You hold the judgment on which target, which message, which moment, and you approve every send. The send layer uses your own account, stays inside the limits, and caps you at 20 actions a day on a new account.

The skills are the easy part. Pointing them is the operator's job. That is the line where Styfinity lives.

FAQ

No. The skills work off data you give them. The send layer is opt-in, runs on your own LinkedIn session or your own provider key, is dry-run until you pass --send, and is hard-capped at 20 actions a day. Nothing happens that you did not approve.

No. All 29 skills run with zero setup. They draft posts, notes, DMs, comments, sequences, and triage off pasted data. The CLI and MCP layer are what turn drafts into a live loop once you are ready.

No, and anyone who promises that is selling you a banned account. This gives you reach and conversations at machine speed, aimed at the right people. The outcomes follow from who you choose and how you show up.

Yes. The send layer targets a generic provider shape (Unipile by default) and every endpoint is overridable in config. See mcp/README.md.

Want the whole engine wired for your pipeline, not just the skills?

That is what we do. I am a Revenue Partner who uses AI to do the heavy lifting. I build the engine, or I run it for you, so the calls land in your calendar without you becoming the operator.

styfinity.com · MIT licensed · fork it, ship it, make it yours

Built by Josh Stylianou, Founder of Styfinity.