Building Your First Claude Skill

# From The Claude Code Playbook — claudecodeguides.com/playbook/


## The Specific Situation

You have a Rust project. Every time you ask Claude Code to explain a module, you paste the same instructions: "show me the ownership flow, highlight unsafe blocks, draw a dependency diagram." You do this three times a day. A skill eliminates that repetition permanently. You create one SKILL.md file, and Claude applies those instructions whenever you ask about code structure -- without you pasting anything.

This works for any language or workflow. The Rust example here targets a real search demand pattern (developers searching "claude rust skills" get zero useful results right now), but the technique applies identically to TypeScript, Python, Go, or anything else.

## Technical Foundation

Claude Code skills follow the Agent Skills open standard (agentskills.io). Each skill is a directory containing a SKILL.md file as its entrypoint. The SKILL.md uses YAML frontmatter for metadata and markdown for instructions.

Skills live in one of three locations:

- **Personal**: `~/.claude/skills/<skill-name>/SKILL.md` -- available in all your projects
- **Project**: `.claude/skills/<skill-name>/SKILL.md` -- shared with your team via git
- **Enterprise**: Managed settings path -- deployed organization-wide

Unlike CLAUDE.md content that loads at session start, a skill's body loads only when invoked. This means long reference material costs almost nothing until you need it. Claude watches skill directories for file changes, so edits take effect immediately without restarting your session.

## The Working SKILL.md

Create the directory and file: