Claude and its skills — a five-section visual primer
Visual primer
Claude, and its skills.md.
Claude shows up across chat, code, design, browser, and desktop. A skill is a folder of your preferences — Claude reads it on demand so the same job comes out the same way every time.
Page 1 of 3 — A skill captures your preferences
01What & why
A skill captures your preferences.
Inside the folder, a SKILL.md says when to use it. Scripts, templates, and references sit alongside. When a request matches, Claude reads the rest — so the same job comes back the same way every time.
Without a skill, the model is smart but generic — it doesn't know your house style, your team's checklist, your formatting rules. With one, it doesn't have to guess. Write down what "good" looks like once, drop the folder in, and every subsequent ask comes back the same way.
Cheaper than retraining. Cheaper than redoing it every time.
Page 2 of 3 — How to actually use a skill
02In practice
Triggered by description. Or by name.
Most of the time, Claude picks the right skill from your phrasing. Type /skill-name when you want to be explicit — same result, more direct.
retention's the real story. Pulling the chart now.
— written in your voice, no preamble, action-first.
Usually the description in SKILL.md matches your ask and Claude loads it on its own. The slash command is for when you want to be sure, or when two skills could fit.
Skills are stored locally. They don't travel between machines unless you copy the folder.
Page 3 of 3 — A few skills people are building
03Examples
A few skills, in the wild.
From personal habits to team-wide playbooks — the same shape, different jobs.
VOICE
SKILL.md. Type /style apply before drafting an email or post and Claude writes the way you do — same rhythm, same hedges, same em-dashes.& SCRAPE
SYSTEM
REVIEW
If you can describe how you do it, you can ship it as a skill.