Where are Claude skills stored?

Personal skills live in `~/.claude/skills/` and project-specific skills in `.claude/skills/` inside the repo. Each skill is its own folder containing a `SKILL.md`.

Do I need to code to create a Claude skill?

No. A skill is plain markdown with a short YAML frontmatter. Scripts are optional. You can also generate one with the free Skill Builder without writing any YAML.

How does Claude decide when to use a skill?

From the `description` field. Claude always sees each skill's name and description, and loads the full body only when the current task matches — so write the description around what the skill does and when it should trigger.

Is a Claude skill the same as an agent skill?

Yes. Claude Skills are Anthropic's implementation of the open Agent Skills / SKILL.md standard, so the same file format works across compatible agents like Codex and Gemini CLI.

How do I share a Claude skill?

Publish the skill folder to a Git repo, or publish it to the public catalog on ProfessorGPT so others can install it with a single command.

How to Create a Claude Skill (SKILL.md)

A Claude skill is a folder with a SKILL.md file that teaches Claude how to do a specific job. This guide walks through the format, a full working example, and how to install it in Claude Code — plus a free generator so you can skip the boilerplate.

What is a Claude skill?

A Claude skill is a reusable capability you give an agent as a small folder containing a SKILL.md file (and optional scripts or reference files). Claude reads the skill's name and description up front, and loads the full instructions only when the current task matches — so skills scale without bloating the context window.

Claude Skills are Anthropic's implementation of the open Agent Skills standard (the same SKILL.md format works in Claude Code, the Claude apps, and other agents). If you're new to the concept, start with what agent skills are.

Claude skill vs command, subagent, plugin, and MCP

These primitives overlap, so pick the right one:

Skill — reusable know-how ("how to review a PR"). Loaded on demand via SKILL.md. Best for repeatable workflows.
Slash command — a manual shortcut you invoke explicitly. A skill triggers itself when relevant; a command doesn't.
Subagent — a separate agent with its own context window for a delegated task. Heavier than a skill.
Plugin — packages commands, skills, and hooks together for distribution.
MCP server — connects the agent to external tools and data. MCP adds *tools*; a skill adds *instructions*. They're complementary.

Anatomy of a SKILL.md file

Every skill needs a SKILL.md with a YAML frontmatter block and a markdown body:

name — lowercase, kebab-case, and it must match the folder name (max 64 chars).
description — one or two sentences stating what the skill does and when it should trigger. This is the single most important field: Claude uses it to decide whether to load the skill.
Body — the actual instructions, written as clear imperative steps. Keep it under ~500 lines; move long references into separate files the skill can point to.

How to create a Claude skill step by step

Follow these steps to build your first skill from scratch.

Installing your skill: the ~/.claude/skills folder

Claude Code discovers personal skills in ~/.claude/skills/ and project skills in .claude/skills/ inside a repo. Create a folder named after your skill and drop the SKILL.md inside:

mkdir -p ~/.claude/skills/code-reviewer
# add ~/.claude/skills/code-reviewer/SKILL.md

Restart Claude Code (or start a new session) and the skill is available — Claude invokes it automatically when a task matches the description. No manual "use this skill" needed.

A complete SKILL.md you can copy

Here is a minimal, working example. Copy it into ~/.claude/skills/code-reviewer/SKILL.md:

---
name: code-reviewer
description: Reviews a diff or pull request for correctness bugs, security issues, and style. Use when the user asks to review code, a diff, or a PR.
---

# Code Reviewer

When asked to review code:

1. Read the full diff before commenting.
2. Flag correctness bugs first, then security issues, then style.
3. Give concrete feedback with file:line references.
4. End with a short summary and a clear approve / request-changes call.

Build a Claude skill without writing YAML

If you'd rather not hand-write frontmatter, the Skill Builder generates a clean, standard-compliant SKILL.md from a few inputs — pick a category, describe the job, and download or install the result. You can also browse ready-made skills and adapt one.