Claude Code skills vs slash commands

Side by side

From the Claude Code, Cursor and GitHub docs, read 5 November 2026.

A worked example: two commands become skills

Illustration: a made-up business. A 7-person product team had two command files: release-notes.md, a page of rules for writing the customer changelog, and deploy.md, which runs the deploy steps.

• Release notes → a skill that loads by itself. Folder .claude/skills/release-notes/, with the description "Writes customer release notes from merged changes. Use when someone asks for release notes, a changelog or 'what shipped'." The style guide moved beside it as a reference file, so the main file stays short.
• Deploy → a manual-only skill. Folder .claude/skills/deploy/, with disable-model-invocation: true, so it runs only when someone types /deploy.

Both still answer to the same /release-notes and /deploy, and the two engineers who use Cursor now get the same release-notes skill, because Cursor also loads .claude/skills. The team tested each by asking in plain words ("what shipped this week?") and checking which one loaded.

Move yours

1. List your command files and mark each "Claude may run it" or "only when I type it".
2. Make a folder per command under .claude/skills/ with a SKILL.md.
3. Write a "use when" description with the words people actually type.
4. Add disable-model-invocation: true to anything with side effects.
5. Move long reference text into files beside SKILL.md; keep the main file under 500 lines.

Next step: share the skills

Skills in one repository reach the people who work in it. To give the same skills to a whole team, see sharing Claude skills on GitHub. For what belongs in AGENTS.md instead, see SKILL.md vs AGENTS.md.

Questions

What is the difference between Claude skills and slash commands?

In Claude Code there's no longer a separate system: custom commands have been merged into skills. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy. A skill adds a folder for supporting files, front matter to control whether you or Claude invokes it, and loading on its own when a request matches its description.

Do my .claude/commands files still work?

Yes. Claude Code's docs say existing command files keep working and the older format is still supported. For new work they recommend a skill, because skills also support supporting files.

How do I stop Claude from running a skill by itself?

Set disable-model-invocation: true in the skill's front matter. Claude then can't invoke it on its own, and you run it with /name. The docs suggest this for workflows with side effects, such as /deploy or /commit.