What Is an AI Skill? Skills, Plugins and Connectors Explained

What is an AI Skill? Learn the difference between Skills, plugins and connectors, and follow a step-by-step guide to building your own Skill in Claude.

As AI assistants like Claude become central to how we work, a new vocabulary has appeared: Skills, plugins, and connectors. They sound similar and are often used interchangeably — but they do different jobs. Understanding the difference helps you build AI workflows that are powerful, reliable, and reusable.

This guide explains each concept clearly and then shows you how to build your own Skill in Claude.

The Big Picture

Think of an AI assistant as a brilliant new team member:

• Connectors give them access to your tools and data (the "where things are").
• Plugins / tools give them new abilities to take actions (the "what they can do").
• Skills teach them how your company does a specific task (the "how we do it here").

You need all three to get the most out of AI — but Skills are the layer that captures your unique know-how.

What Is a Connector?

A connector links an AI assistant to an external system or data source — your email, calendar, CRM, Google Drive, databases, or third-party apps.

Connectors answer the question: "What can the AI see and reach?"

Examples:

• A Google Drive connector lets Claude read your documents.
• A CRM connector lets it look up customer records.
• A database connector lets it query live business data.

Connectors are about access and integration.

What Is a Plugin / Tool?

A plugin (often called a tool or function) gives the AI a new capability — an action it can perform. Where a connector provides access, a tool provides ability.

Examples:

• A tool that runs a calculation.
• A tool that creates a calendar event.
• A tool that sends a message or files a ticket.
• A tool that searches the web.

Plugins are about doing things — taking action beyond generating text.

What Is a Skill?

A Skill is a reusable package of instructions, knowledge, and optional scripts that teaches the AI how to perform a specific task the way you want it done.

A connector says where the data is. A tool says what the AI can do. A **Skill says how to do a job properly** — your steps, your standards, your templates.

Skills answer: "How does our company do this particular task?"

Examples:

• A Skill that writes proposals in your exact format.
• A Skill that produces reports using your KPIs and template.
• A Skill that reviews contracts against your checklist.
• A Skill that formats data into a branded slide deck.

Crucially, Skills load only when relevant. The AI reads a short description of each Skill and pulls in the full instructions when the task matches — keeping it focused and efficient.

How They Work Together

``text ┌──────────────────────────────┐ │ AI Assistant │ └──────────────────────────────┘ │ │ │ Connectors Plugins Skills (access) (actions) (know-how) │ │ │ Drive, CRM, Send email, "Write our database run script monthly report" ``

A real task might use all three: a connector pulls the data, a Skill defines how to analyse and format it, and a plugin delivers the finished output to the right place.

How to Build a Skill in Claude

A Skill is essentially a folder with a main instruction file and optional supporting files. Here is the process.

Step 1: Pick one specific, repeatable task

Skills work best when narrowly scoped. "Write our weekly client status report" is a great Skill. "Help with marketing" is too broad.

Step 2: Create the folder structure

``text weekly-status-report/ ├── SKILL.md # required: instructions + when to use ├── references/ │ └── definitions.md # optional: terms, KPI definitions └── assets/ └── template.md # optional: the report template ``

Step 3: Write the SKILL.md

Every Skill needs a name, a description, and instructions. The description is the most important line — it is how the AI decides when to use the Skill.

```markdown --- name: weekly-status-report description: Use when asked to write a weekly client status report. Produces a report in our standard format using the project data provided. ---

Weekly Status Report

When to use

Triggered when the user asks for a weekly client or project status report.

Instructions

1. Read the project data provided.
2. Follow the template in assets/template.md exactly.
3. Summarise progress, blockers, and next steps.
4. Keep the tone professional and concise.
5. Flag any item that is overdue.

Reference

• See references/definitions.md for how we define "on track" vs "at risk".

```

Step 4: Add references and assets

Put templates, examples, and definitions in the supporting files. This keeps the main instructions short while giving the AI everything it needs.

Step 5: Add scripts (optional)

For repeatable processing — like reformatting data or generating a file — you can include small scripts the Skill runs.

Step 6: Test and refine

Run the Skill on real examples. Check that it triggers at the right time and produces output to your standard. Tighten the description and instructions based on results.

Best Practices

• One job per Skill. Narrow scope beats a do-everything Skill.
• Nail the description. It controls when the Skill activates.
• Keep instructions clear and ordered. Numbered steps work well.
• Use references for detail. Keep the main file lean.
• Iterate. Treat Skills as living assets you improve over time.

Why This Matters for Business

Skills turn institutional knowledge into reusable AI capability. Instead of relying on individuals to remember "how we do it," you encode it once — and every team member gets consistent, high-quality output.

At Adaptive Media we help businesses design connectors, tools, and custom Skills that make AI genuinely useful for the way you work. If you want to build a Skill library that captures your company's expertise, we can help you plan and build it.