Claude Code Skills: The Complete Guide (With 8 Business Examples)

Quick answer: Claude Code Skills are reusable instruction folders (a SKILL.md file plus optional templates and scripts) that teach Claude how to do a specific job your way. Claude loads a skill only when it's needed, either automatically or when you type /skill-name, so your best processes run the same way every time.

Most teams use AI the same way every day: open a chat, paste the same long instructions, fix the same mistakes, repeat. It works, but it doesn't scale.

Claude Code Skills fix that. A skill is a saved, reusable set of instructions that Claude loads only when it's needed. Write your process once (how you format a client report, how you reply to a support ticket, how you review code) and Claude follows it every time, with a single command or automatically when the task comes up.

This guide covers what Claude Code skills are, how the SKILL.md file works, which skills come built in, how to install skills from a marketplace or GitHub, how to build your first skill step by step, and 8 practical ways businesses can use them.

What Are Claude Code Skills?

A Claude Code skill is a folder that contains a SKILL.md file. That file has two parts:

1. A short header (frontmatter) that says what the skill does and when to use it.
2. Instructions in plain Markdown: the steps, rules and examples Claude should follow.

A skill can also include extra files: templates, reference documents, examples or scripts.

You can run a skill yourself by typing /skill-name, or Claude can pick it up automatically when your request matches the skill's description. Ask "can you write this week's client report?" and Claude can load your weekly-client-report skill without being told.

Skills follow the Agent Skills open standard, so the same skill format also works in other Claude products, such as Claude on the web and Claude Cowork, when a skill is enabled for your account.

In one line: CLAUDE.md tells Claude who you are; a skill teaches Claude how to do a specific job, and it only loads when that job comes up.

How Claude Code Skills Work (and Why They're Efficient)

Skills use something called progressive disclosure. Claude doesn't read every skill in full at the start of a session. Instead:

1. Always loaded: only each skill's short description, so Claude knows what's available.
2. Loaded when used: the full SKILL.md instructions, only when you run the skill or Claude decides it's relevant.
3. Loaded only if needed: supporting files (like reference.md or a template), when the instructions point to them. Scripts are run, not read, so their code doesn't fill up Claude's memory.

This means you can have dozens of skills with long reference material, and they cost almost nothing until they're actually used.

Where Claude Code Skills Live

Where you save a skill decides who can use it:

Rule of thumb: personal habits go in your personal folder; team processes go in the project folder so the whole team gets the same skill through Git.

Claude Code Skills vs CLAUDE.md vs MCP vs Subagents

These four features are often confused. Here's how they differ:

They work best together. For example, an MCP server connects Claude to your CRM, and a skill tells Claude exactly how your team wants leads scored and notes written.

What about slash commands? Custom commands have been merged into skills. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy. Old command files keep working, but skills add supporting files, automatic loading and more control.

Why Businesses Should Care About Claude Code Skills

Skills aren't just a developer feature. Any team that repeats the same AI tasks benefits:

• Consistent output. Every report, reply and summary follows the same structure, no matter who on the team runs it.
• Less prompting. Nobody has to remember or paste the "perfect prompt". It lives in the skill.
• Faster onboarding. New team members get your best process on day one by running /skill-name.
• Process knowledge stays with the company. Your way of working is written down, version-controlled and easy to improve.
• Lower cost. Because skills load only when needed, long instructions don't waste tokens on every message.

Built-in Skills That Come With Claude Code

You don't have to start from zero. Claude Code ships with bundled skills you can use right away by typing / and the skill name:

The exact list grows with each Claude Code release, so type / to see what's available in your version.

How to Install Claude Code Skills (Marketplace or GitHub)

Besides writing your own, you can install skills other people have built. There are two common ways.

Option 1: Install From a Plugin Marketplace (Recommended)

Skills are often shared inside plugins. A plugin can bundle several skills, plus agents, hooks and tool connections. Anthropic runs an official marketplace that Claude Code adds for you automatically.

1. In Claude Code, type /plugin to open the plugin browser and search the Discover tab.
2. Or install a plugin directly by name, for example Anthropic's skill-creator, which helps you write new skills:

/plugin install skill-creator@claude-plugins-official

3. Choose who gets it: just you, everyone on the project, or you in this project only.
4. Type / and the plugin's skills appear as /plugin-name:skill-name.

To install from another marketplace hosted on GitHub, add it first:

/plugin marketplace add owner/repo

In the Claude desktop app: click the + button next to the prompt box and choose Add plugins.

Option 2: Copy a Skill Folder From GitHub

Many open-source skill collections live on GitHub. To use one skill from them, copy its folder (the one containing SKILL.md) into ~/.claude/skills/ for yourself, or .claude/skills/ in your project for your team. Claude Code picks it up automatically.

Where to Find Claude Code Skills

• Anthropic's official marketplace: browse it with /plugin in Claude Code or on claude.com/marketplace.
• Community GitHub collections: open-source libraries with hundreds of skills for coding, writing, design and more.
• Your own team: the most valuable skills are usually the ones that capture your processes.

Before installing any skill or plugin, read its SKILL.md and any scripts. Skills and plugins can run commands on your machine, so only install from sources you trust.

How to Create Your First Claude Code Skill (Step by Step)

Let's build a real skill: a weekly client report generator.

Step 1: Create the Skill Folder

For a skill you want in every project, create a folder in your personal skills directory:

mkdir -p ~/.claude/skills/weekly-client-report

For a team skill, use .claude/skills/weekly-client-report/ inside your project instead.

Step 2: Write the SKILL.md File

Create SKILL.md inside that folder:

---
name: weekly-client-report
description: Builds the weekly client performance report from a CSV export. Use when the user asks for a weekly report, client update or performance summary.
argument-hint: [client-name] [csv-file]
disable-model-invocation: true
---

# Weekly client report

Create the weekly report for client $0 using the data in $1.

## Steps
1. Read the CSV and calculate total leads, conversion rate and week-over-week change.
2. Fill in the template in [template.md](template.md). Keep the section order.
3. Write a 3-bullet "What changed" summary in plain English.
4. List any metric that dropped more than 20% under "Needs attention".
5. Save the report as reports/$0-weekly-report.md.

## Rules
- Never invent numbers. If a column is missing, say so.
- Use the client's name exactly as written.

What each header field does:

• name: the command you type (/weekly-client-report). If left out, the folder name is used.
• description: the most important line. Claude reads it to decide when the skill is relevant. Put the main use case first and include the phrases people actually say.
• argument-hint: shows what to type after the command.
• disable-model-invocation: true: only you can start this skill. Use it for anything that creates files, sends messages or deploys code.

$0 and $1 are replaced with the first and second words you type after the command.

Step 3: Add Supporting Files (Optional)

Add a template.md with your report layout next to SKILL.md. Claude only opens it when the instructions say to.

weekly-client-report/
├── SKILL.md        (required: instructions)
├── template.md     (report layout)
├── examples.md     (a good past report)
└── scripts/
    └── calculate.py  (run, not read)

Step 4: Run It

In Claude Code, type:

/weekly-client-report acme-co data/acme-week-43.csv

Claude follows your steps and saves the finished report.

Step 5: Test and Improve

Test the skill in a fresh session, both by calling it directly and by asking naturally ("can you put together Acme's weekly update?"). If Claude doesn't pick it up automatically, improve the description. If the output is off, tighten the steps or add an example.

8 Claude Code Skills Examples for Business Automation

Example use cases: illustrative workflows showing how teams can use skills.

1. Weekly Client Report Generator

Turns a CSV or spreadsheet export into a formatted client report with highlights, drops and next steps. Saves: 1–2 hours per client, per week.

2. Invoice-to-Accounting Formatter

Reads supplier invoices (PDF or email text), pulls out vendor, date, line items and tax, and outputs a clean import file for QuickBooks or Xero. A reference file holds your chart of accounts so Claude codes expenses the same way every time.

3. Brand-Voice Content Writer

Stores your brand guidelines, banned words and example posts in supporting files. Anyone on the team can run /brand-post and get LinkedIn posts, emails or product descriptions that sound like your company, not like generic AI.

4. CRM Lead Enrichment and Notes

Combined with an MCP connection to your CRM (like HubSpot), the skill defines how to score a lead, which fields to fill and how to write the call-prep note. The MCP gives access; the skill gives the process.

5. Support Reply Drafter

Loads your help-center articles, refund policy and tone rules only when a support ticket comes up. It drafts the reply, flags anything that needs a human (refunds above a limit, angry customers, legal questions) and never promises what the policy doesn't allow.

6. SEO Blog Brief Generator

Give it a keyword and it produces a full content brief: search intent, outline, questions to answer, internal links and a meta description, in your team's exact brief template.

7. Workflow Documentation Writer

Point it at an exported automation (for example an n8n or Make workflow file) and it writes a plain-English SOP: what triggers it, what each step does, what can fail and who owns it. Great for handovers.

8. Pre-Deploy Code Review Checklist

A developer-side skill set to manual-only (disable-model-invocation: true). Running /predeploy checks for missing tests, hard-coded secrets, unhandled errors and migration risks before anything ships.

Best Practices for Writing Claude Code Skills

• Write a sharp description. Start with the main use case and include trigger phrases ("weekly report", "client update"). The description and when_to_use text together are cut off at 1,536 characters in the skill list, so keep the key use case first.
• Keep SKILL.md short. Aim for under 500 lines. Move long reference material into separate files and link to them.
• Give instructions, not essays. "Do X, then Y" works better than explaining why.
• Separate knowledge from tasks. Reference skills (brand rules, conventions) can load automatically. Task skills that change things (send, deploy, delete) should use disable-model-invocation: true.
• Include one good example. A sample of the finished output is often worth more than ten rules.
• Test in a fresh session. Your own context can hide gaps that a teammate would hit.
• Version it. Keep team skills in Git so improvements are shared and reviewed.

Common Mistakes to Avoid

1. A vague description like "helps with reports". Claude won't know when to use it.
2. One giant skill that tries to do everything. Split it into focused skills.
3. Letting Claude auto-run risky actions. Anything with side effects should be manual-only.
4. Hard-coding secrets (API keys, passwords) inside skill files.
5. Never updating. Treat skills like documentation and review them when your process changes.

Are Claude Code Skills Safe?

Skills can include scripts and shell commands, so treat them like code:

• Only install skills from sources you trust, and read the SKILL.md and any scripts first.
• Check allowed-tools in shared skills. This field lets a skill use certain tools without asking for permission each time.
• Companies can lock things down with managed settings, for example blocking shell commands inside skills.

How BaruzoTech Helps Teams Use Claude Code Skills

Writing one skill is easy. Building a system of skills that fits your real operations, connected to your CRM, inbox, accounting and internal tools, is where most teams get stuck.

At BaruzoTech, we design and build AI systems that run your operations. That includes:

• Mapping which repeatable tasks are worth turning into skills
• Building skills, MCP connections and automations that work together
• Adding guardrails: approvals, logs and human-in-the-loop steps for sensitive actions
• Supporting your team as processes change

Want to know which of your team's tasks could become one-command workflows? Book a free systems call → You'll get an honest take on what's worth automating.

Frequently Asked Questions

What are Claude Code Skills?

Claude Code Skills are folders containing a SKILL.md file with instructions that teach Claude how to do a specific task. Claude loads a skill automatically when it's relevant, or you run it with /skill-name.

Where are Claude Code skills stored?

Personal skills go in ~/.claude/skills/<skill-name>/SKILL.md and work across all your projects. Project skills go in .claude/skills/<skill-name>/SKILL.md inside a repository and can be shared with your team through Git.

What's the difference between Claude Code skills and slash commands?

Custom slash commands have been merged into skills. Both create a /command, but skills add supporting files, automatic loading and settings that control who can run them.

What is the difference between Claude skills and MCP?

MCP connects Claude to outside tools and data, like a CRM or database. Skills teach Claude how to do a task. They work best together: MCP for access, skills for the process.

Do Claude skills work outside Claude Code?

Yes. Skills follow the open Agent Skills standard. Skills enabled on your Claude account can be used in Claude on the web and in Claude Cowork, although some Claude Code–only settings aren't supported there.

What skills come built into Claude Code?

Claude Code includes bundled skills such as /doctor, /code-review, /simplify, /loop, /claude-api, /run, /verify and /skill-doctor. Type / in Claude Code to see the full list for your version.

How do I install a Claude Code skill from GitHub?

Either add the GitHub repository as a marketplace with /plugin marketplace add owner/repo and install the plugin, or copy the skill's folder (containing SKILL.md) into ~/.claude/skills/ or your project's .claude/skills/ folder.

Where can I find Claude Code skills?

In Anthropic's official plugin marketplace (open it with /plugin in Claude Code), in open-source skill collections on GitHub, or by writing your own for your team's processes. Always review a skill before installing it.

How long should a SKILL.md file be?

Keep it under about 500 lines. Put long reference material into separate files that the skill links to, so they only load when needed.

Do I need to be a developer to create a skill?

No. A skill is mostly plain-English Markdown. Developers help when a skill needs scripts or tool connections, but the instructions themselves can be written by anyone who knows the process.

Related reading: What is Claude Cowork? · WhatsApp business automation: 10 use cases