Claude Code Skills: Reusable Expertise

TL;DR

• Claude Code skills are reusable expertise files that tell Claude exactly how to do a specific task.
• Each skill lives in a SKILL.md file and gets picked up automatically when Claude Code starts.
• They stop you from re-explaining the same process every single session.
• Skills work across projects, teams, and agents so your best thinking scales.
• You write a skill once, then it works forever.

What are Claude Code skills?

Claude Code skills are the answer to a problem every heavy user hits. You explain a process once. It works great. Next session, you explain it again. Same thing. Over and over.

A skill is a SKILL.md file that captures a repeatable process in plain text. Claude Code finds it automatically. You never re-explain that process again.

Each SKILL.md file tells Claude:

• What the skill does
• When to use it
• The exact steps to follow
• Any rules or edge cases to watch for

That is it. No code required. No plugins. Just a markdown file with the right name in the right place. If you are new to Claude Code, start with the complete guide to Claude Code before diving into skills.

Why does reusability matter?

Here is the problem with AI without skills. Every session starts cold. Claude knows nothing about your business, your standards, or your preferred way of doing things. You spend the first ten minutes re-establishing context.

That is fine for a one-off task. For recurring work, it kills productivity.

Think about what happens at Devwiz or across the 14+ clients on the Njin Method. The same types of tasks come up constantly. Blog posts. Client briefs. Code reviews. Status updates. Without skills, every agent and every session handles these differently.

With skills, the process is locked in. The output is consistent. The quality stays high whether it is the first time or the hundredth.

Skills are how you stop being the bottleneck in your own AI system.

How do skills actually work?

Claude Code has auto-discovery built in. When a session starts, it scans for SKILL.md files in a set of known locations.

The main locations are a .claude/skills/ folder inside your project, and a .claude/skills/ folder in your home directory for global skills. A skill in your project folder is available only in that project. A skill in your home directory is available everywhere.

When Claude needs to do something and a skill exists for it, it follows the skill. No prompting needed. You can also call a skill directly by name if you want to run it on demand. The file must be named SKILL.md, uppercase. That is not optional. Claude Code looks for exactly that name.

Skills work alongside Claude Code memory and hooks to build a full system that runs the way you want it to run.

What goes inside a SKILL.md file?

The structure is flexible but a good skill usually has these sections:

Name and purpose. One or two sentences on what this skill does and when it applies.

Trigger conditions. How does Claude know to use this skill? Is it automatic? Does the user have to ask for it?

Step-by-step process. The actual instructions. Write these the same way you would explain the task to a new hire. Clear, sequential, no assumptions.

Rules and standards. Any constraints that must hold. Voice guidelines, file naming rules, output formats, things to never do.

Examples. Optional but useful. A sample input and output shows Claude exactly what done looks like.

Keep it focused. One skill, one process. If a skill is trying to do three things, split it into three skills.

Real examples from daily use

Here are skills I use constantly across Njin Method client work and my own projects.

Blog post skill. This skill covers every step of writing a blog post. It sets the structure, the voice rules, the SEO requirements, the FAQ format, the TL;DR format, and the reading time calculation. Every blog post comes out consistent. The skill took about an hour to write and has saved hundreds of hours since.

Voice humaniser skill. Any AI-drafted content goes through this skill before it gets used. The skill runs a compliance check against a banned-words list, checks for Australian English, checks reading level, and rewrites anything that fails. No more proofreading for the same issues every time.

Session archive skill. At the end of a session, this skill writes a summary to a standard location. It captures what was done, what decisions were made, and what comes next. Context never gets lost between sessions.

Client brief skill. For Njin Method onboarding, this skill pulls the right questions, formats the output into a standard brief, and flags anything missing. Every client starts with the same quality of brief.

Each of these replaces a process I used to explain from scratch. Now they just work. Skills also connect well with subagents when you want different agents to share the same expertise without re-teaching each one.

How to build your first skill

Pick one process you repeat at least once a week. That is your first skill.

Open a text editor. Create a file called SKILL.md. Write down:

1. What this process produces
2. When to run it
3. Every step in order
4. Any rules that must hold

Save it to the .claude/skills/ folder inside the project where you use it most. Start a new Claude Code session and test it.

Expect to revise it. The first version will miss things. Run the process a few times, notice where Claude goes off-track, and update the skill. After two or three iterations, it will be solid. Then move it to your home .claude/skills/ folder if you want it available globally.

One good skill beats ten clever prompts.

Related reading

• Claude Skills explained: knowledge vs how-to Skills beyond the terminal.
• Skills vs Projects vs Subagents vs MCP which to reach for, and when.

---

FAQ

What is the difference between a skill and a prompt?
A prompt is something you type once for one task. A skill is a stored process that runs the same way every time, automatically, without you typing anything. Skills are reusable. Prompts are not.

Can I share skills across projects?
Yes. Skills in your home .claude/skills/ folder are available in every project. Skills inside a project folder are local to that project. Put skills in the global folder when multiple projects need the same process.

How many skills should I have?
Start with three to five covering your most repeated tasks. Add more as you identify gaps. There is no hard limit, but skills that overlap or conflict will cause confusion. Keep them focused.

Do skills work with Claude Code agents?
Yes. When you spawn a subagent, it can access the same skills as the parent session. This is how you get consistent behaviour across a multi-agent system without re-teaching each agent.

---

If you want to build a proper AI system around your business, skills are one of the first things to get right. They are the mechanism that turns one good process into a repeatable one. Book a call and let us get started.