Claude Skills

Claude Skills extend what Claude Code can do. Official docs say you create a SKILL.md with instructions and Claude adds it to its toolkit. Claude can load a skill when it is relevant, or you invoke it with /skill-name. Use a skill when you keep pasting the same checklist into chat, or when a CLAUDE.md section has grown into a procedure. Unlike CLAUDE.md, the skill body loads only when used.

Custom slash commands have been merged into skills. .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md both create /deploy. Existing command files still work. Skills add a directory for supporting files, frontmatter that controls who invokes them, and automatic loading. Claude Code follows the Agent Skills standard and adds invocation control, subagent execution, and dynamic context injection.

Key Features

• SKILL.md packages: Instructions plus optional supporting files, not a one-off prompt.
• Slash invocation: /skill-name from the CLI, plus auto-load when relevant.
• Commands merged in: Old .claude/commands/ paths keep working.
• Open standard: Portable Agent Skills, with Claude-only extras on top.
• Can run in a subagent: Docs describe running a skill in an isolated worker so the main thread stays clean.

Use Cases

• Repeatable repo chores: deploy, release, review checklists.
• Team procedures you do not want in every session's CLAUDE.md.
• Cross-tool skills that should also work in other Agent Skills clients.

Pricing

Skills ship inside Claude Code. There is no separate skills SKU. You pay the Claude subscription or Console/API usage that already runs Claude Code. Recheck claude.com/pricing.

Getting Started

1. Read Extend Claude with skills.
2. Add .claude/skills/<name>/SKILL.md in the repo (or a user-level skill directory from the same docs).
3. Invoke with /name once, then let Claude auto-load on the next matching task.
4. For bundled /debug and /code-review, use the commands reference instead of reinventing them.

Frequently Asked Questions

Are skills the same as CLAUDE.md?

No. CLAUDE.md is always-on project memory. A skill loads when used, so long procedures stay cheap.

What happened to custom commands?

They merged into skills. Both file layouts create the same slash command.

Do I need a plugin?

Not for a personal or repo skill. Package skills into a plugin when you want to share a versioned bundle.

Alternatives

• Claude Hooks: Deterministic shell at lifecycle events, not LLM instructions.
• Claude Plugins: How you ship skills to a team.
• Claude Code: The host product.

Tips

1. Move pasted checklists out of CLAUDE.md and into a skill.
2. Keep secrets out of SKILL.md. Put them in hooks or env.
3. Recheck the skills doc when Anthropic changes invocation frontmatter.

Conclusion

Skills are the current way to extend Claude Code with reusable procedures. Start from the official skills guide, write one SKILL.md, and only then package it as a plugin.