Expandi Skills

Expandi Skills

Teach the AI you already use to run LinkedIn outreach like someone who has done it before. Each skill covers one job, from choosing who to reach to reading the replies, and most of them work before you connect an Expandi account. You brief the agent, it does the work, and you approve what goes out.

Browse the library See how it works
Getting started

How it works

  1. Connect the MCP

    Link your AI platform to Expandi MCP (the steps for each platform are in the Help Center). Most skills work without an Expandi account; the platform picker below says which ones need it.

  2. Download a skill

    Download the skill you need from its page below. One quick form, and you’ll hear from us whenever that specific skill ships an update.

  3. Brief your agent

    Talk to your agent the way you’d brief a teammate, describe what you’re trying to do, and the right skill activates on its own.

The library

Browse by category

Plan your outreach

More planning skills, a full plan-before-you-spend-a-connection-request builder, and an agency/client-facing strategy doc, are coming soon.

Write messages that land

More copy skills, InMail subject/opener testing, a personalization planner, and a live-campaign copy doctor, are coming soon.

Find people to contact

More sourcing skills, free LinkedIn search, post-engagement mining, event and group attendee campaigns, company followers, and warm-pool revival, are coming soon.

Launch campaigns

More launch skills, cloning a proven template to a new seat, and safely pausing/resuming/transferring leads, are coming soon.

Manage your inbox

More inbox skills, a pre-call lead briefing card and a CRM-ready interested-lead handoff, are coming soon.

Read your numbers

More reporting skills, a Monday multi-seat digest and a statistically honest A/B verdict, are coming soon.

Learn from your results

More learning skills, the compounding learnings ledger, pre-registered next-experiment design, and a bootstrap for everything you’ve already run, are coming soon.

Keep your account safe

A runway guard that projects time-to-empty per campaign, with headroom rails on any top-up, is coming soon.

Good to know

Installing more than one skill?

Install only what the job in front of you needs. If you want to set up a whole role at once, hold off for suites: pre-bundled sets of eight skills or fewer, matched to a job like SDR or agency ops, coming in a later release.

That cap isn’t arbitrary. Some agents, Codex in particular, read every installed skill’s description on startup to work out which one to use, and that lookup runs against a limited budget.

All 15 launch skills still fit comfortably today. As the library grows, installing everything at once is how that budget gets crowded out. So suites, not “download all of it,” will be the recommended way to install more than a couple of skills at a time.

Where it runs

Platform requirements

claude.ai

Read-only skills
Works on any plan
Skills that write to Expandi
Needs a paid plan with a custom MCP connector
Setup
Settings → Capabilities → enable code execution → Skills → upload the zip
FAQ

Frequently asked questions

What is a skill, exactly?

An AI-agent skill is a small folder, one SKILL.md file plus, sometimes, a few reference files, that teaches your AI assistant exactly how to do one job well, in the open Agent Skills format. Install it once and your agent picks it up on its own whenever it’s relevant, instead of you re-explaining the same context in every conversation. Every skill here either operates Expandi directly through its official MCP server, or teaches the LinkedIn outreach craft around it: targeting, copywriting and campaign analysis. Most of them deliver full value with no Expandi account at all.

Which platforms does this actually work on?

Claude.ai, Claude Code, ChatGPT desktop, Codex, GitHub Copilot, M365 Copilot, and Cursor, one canonical skill folder, no per-platform rewrites. See the table above for the exact install step on each. On ChatGPT, a downloadable skill installs in the desktop app, Codex CLI, or the IDE extension; ChatGPT on the web and on mobile runs only the skills a plugin bundles with it, and this library ships standalone downloads, so use the example prompts on each skill page there instead. Skills that write to your Expandi account, like drafting a campaign or queuing a reply, need the Expandi MCP connected and run under your workspace’s tool-approval settings, and every write in these skills asks you first regardless.

Is it safe to let an agent touch my Expandi account?

Every skill that can write to Expandi follows the same rules. It never activates a campaign, the most it ever does is produce a draft, and turning a draft on is always something you do by hand inside Expandi. It never sends a message on its own either, replies are drafted, shown to you exactly as they’d go out, and only queued after you say yes, with an explicit warning first that a queued message can’t be pulled back. Anything destructive, like deleting a lead, is confirmed with the exact undo path shown before it happens. No skill ever asks you for a password or an API key in the chat: your Expandi connection is set up once, on your own platform, through the official Expandi MCP, and every write action a skill can take is drafts-only and approval-gated, so you always see what it would do before it does it.

How do I know if a skill I installed is out of date?

Every skill points at expandi.io/skills/versions.json. A small public file listing the current version of every skill and the shared safety rules they all carry. If a skill ever misbehaves, it’s written to tell you to check that file for a newer release yourself. If you filled out the download form for a skill, you’re also on that skill’s update list: ship a new version, and you get one short note with a fresh download link. The same link always serves the current file, so an old link never goes stale.