Claude now shows skills, connectors and plugins side by side under one Customize menu, and the difference is not obvious. Here is how I explain it to the working professionals I mentor.
The simple version
Think about training a new colleague.
- A skill is the SOP you hand them. It is written instructions for one task, done your way. Your weekly report format. Your client follow-up email. Your checklist for reviewing a contract.
- A connector is the access you give them. Their login to Gmail, Google Drive, Slack or your calendar, so they can actually find things and act.
- A plugin is the full onboarding kit. The SOPs and the logins packaged together, so everything is set up in one go.
Anthropic describes it the same way: connectors give Claude access to your tools, skills teach it how you do a task, and plugins package both as one install.
Skills: teach Claude your way of working
A skill is a folder with an instruction file (called SKILL.md), and sometimes templates, reference files or scripts. Claude reads each skill's short description and loads the full instructions only when your task needs it.
That is the big difference from a long prompt you paste every time, and from a Project:
- Projects hold background knowledge that is always loaded inside that project.
- Custom instructions apply to every conversation.
- Skills switch on only for the task they are made for, and work across your chats.
Claude also ships with built-in skills for Excel, Word, PowerPoint and PDF files. If you ask for a presentation, it uses the PowerPoint skill on its own.
Connectors: give Claude access to your apps
A connector lets Claude read from an app and take actions in it. Search your Drive, summarise your inbox, create a task, send a message.
Two things I always tell learners:
- Claude only sees what you can see. Anthropic says a connector inherits your permissions in that app. If you cannot open a file, Claude cannot either.
- Your work laptop may say no. Company-managed accounts often need IT approval for connectors. Check your company's AI policy first, and use the tools your employer has approved, such as Microsoft Copilot, for work data.
Plugins: everything in one install
A plugin bundles skills, connectors, commands and helper agents into one package. When a plugin exists for a product you use, it comes with that product's connector plus the skills that use it. You add the plugin, sign in to its connector, and you are ready.
Plugins are for paid plans. Skills and connectors are not limited to paid plans.
Which one do you actually need?
| You want Claude to... | Use |
|---|---|
| Follow your format or process every time | A skill |
| Read your email, files or calendar | A connector |
| Get a ready-made setup for a tool or a team | A plugin |
For most of my learners, the answer is: one or two skills and the connectors they need. Plugins come later, usually when a team wants everyone on the same setup.
Build your first skill
Here is a real example from a session. A learner wanted a daily money check. We connected Gmail and asked Claude to create a skill that looks at apartment association notices, bank transaction alerts and home-loan updates, ignores bank promotions, sends a short summary by email, and stays quiet on days with nothing new.
You can do the same with any task you repeat:
- Pick one boring, repeated task. A weekly update, meeting notes into action items, a client follow-up.
- Connect the app it needs. Go to Customize, then Connectors.
- Ask Claude to build the skill with you. Say "I want to create a skill for..." and answer its questions. Share a template or a good past example if you have one.
- Save it and switch it on in Customize, then Skills.
- Test it in a new chat. Watch for permission requests or errors before you rely on it.
- Fix it by talking to Claude. If the output is off, ask Claude to update the skill.
A tip from my sessions: make separate skills for separate outputs. A client presentation and an internal presentation deserve two skills, not one.
Stay safe
Anthropic warns that a skill can include instructions or code that make Claude do things you did not intend. Only add skills from sources you trust, read what is inside one before you turn it on, and see my guides on keeping your data private in ChatGPT, Claude and Gemini and Claude's auto mode.