Guide

How to use Claude's skill-creator to draft and test a skill

skill-creator is a skill from Anthropic that helps you write other skills. Turn it on under Customize > Skills in claude.ai or the desktop app, where it's listed as from Anthropic, then ask Claude to use it. It drafts the skill with you from a task you describe, then runs the skill on test prompts you agree on, shows the results and helps you revise. It does more in Cowork and Claude Code than in chat: there it also runs the same prompts without your skill as a baseline, in parallel, with pass rates, timing and token counts, so you can see whether the skill actually improves the output.

Short answer

  • Turn it on: Customize > Skills, from Anthropic.
  • Ask for it: "use skill-creator to make a skill for…".
  • Chat: tests one prompt at a time.
  • Cowork and Claude Code: adds a baseline without the skill, pass rates and timing.

Before you start

  1. Check the plan: Anthropic's Help Center lists skills on Free, Pro, Max, Ultra, Team and Enterprise; the Claude docs overview lists Pro and up. Sharing a skill with colleagues needs Team or Enterprise.
  2. Turn on Code execution and file creation under Settings > Capabilities (on Team and Enterprise, an Owner does it for the organization).
  3. Write the task in one sentence and what a good result looks like. That sentence becomes the skill's description.
  4. Collect three real inputs to test with, including one awkward one.

A worked example

Illustration: a made-up business. Ama runs support at a 15-person company on Claude Team and corrects the same things in every refund reply. She turns skill-creator on and writes: "Use skill-creator to make a skill for refund replies: our policy, our tone, and a reply under 120 words that says what happens next." It asks her for the policy and two replies she liked, then drafts:

  • Name: refund-reply, matching the folder.
  • Description: "Writes a refund reply in our policy and tone. Use when a customer asks for a refund, a return or money back."
  • Steps: check the order date against the 30-day policy; say yes, no or partial in the first line; one next step with a date; under 120 words. Plus her correction: "never promise a refund before finance approves anything over the limit".
  • One example and four checks: the decision is in the first line; the date is right; the word count holds; no promise above the limit.

In Cowork she asks it to evaluate the skill on five refund emails. It runs them with and without the skill: without it, two replies promised refunds above the limit; with it, none did. She keeps the skill, then shares it with her team from Customize > Skills.

Package and check it

  1. Keep the folder name the same as the name in SKILL.md.
  2. Zip the folder itself, so the archive holds refund-reply/SKILL.md, not a loose SKILL.md.
  3. Keep SKILL.md under 500 lines; move long reference material into files beside it.
  4. Read the file it hands you. On claude.ai, skill-creator can package the skill as a .skill file, a ZIP of the folder, for you to download. Open it in the free .skill file viewer to read every file and check the name, description and folder before you upload it.
  5. Upload it under Customize > Skills and turn it on.
  6. Share or publish it on Team and Enterprise with the Share and Publish to org controls.

Next step: share it without a pile of copies

A skill in one account helps one person; shared copies drift. The cheapest way to share AI prompts with a small team compares the options, and skills in Claude Cowork covers where Cowork finds them.

Skills for this

All skills

Questions

What is Claude's skill-creator?

A skill from Anthropic that helps you write other skills. Ask Claude to use it, and it drafts a skill with you, then runs your skill on test prompts you agree on, shows the results and helps you revise it. On claude.ai you turn it on under Customize > Skills, where it's listed as from Anthropic.

Does skill-creator work the same in chat and Cowork?

No. Anthropic's docs say that in chat it works through the test prompts one at a time. In Cowork and Claude Code it also runs the same prompts without your skill as a baseline, runs everything in parallel, and adds pass rates, timing and token counts.

What do I need to use skills in Claude?

Anthropic's two pages differ on plans: the Help Center (updated this week) says skills are available on Free, Pro, Max, Ultra, Team and Enterprise, while the Claude docs overview lists Pro, Max, Ultra, Team and Enterprise. Either way, code execution must be on, under Settings > Capabilities; on Team and Enterprise, an Owner turns it on for the organization.

AI skill

Have your AI turn a prompt that works into a skill your team can reuse

Put this skill in AI Skill Pod and everyone on your team gets it in their Claude, Cursor or ChatGPT (Pro, Business, Enterprise): one invite link, free to start. Turn a prompt into a skill has your AI turn any task you repeat into a skill: a clear name, a "use when" description, the steps and corrections you keep giving, a template, one example and checks, ready to test with skill-creator.

Use this skill in your AI

14 days free · No card needed

Sources

  1. Claude Docs: Create custom skills — ask Claude to use the skill-creator skill to draft a skill; turn skill-creator on under Customize > Skills, listed as from Anthropic; in chat it works through test prompts one at a time; in Cowork and Claude Code it also runs a baseline without the skill, in parallel, with pass rates, timing and token counts; ZIP the skill folder itself; name matches the folder; SKILL.md under 500 lines; Share and Publish to org on Team and Enterprise (read 2026-11-05)
  2. Claude Docs: Skills overview — Pro, Max, Ultra, Team and Enterprise; Code execution and file creation must be on; Claude reads only names and descriptions until a request matches (read 2026-11-05)
  3. Claude Help Center: Use skills in Claude ("Updated this week") — "Skills are available for users on Free, Pro, Max, Ultra, Team, and Enterprise plans"; "This feature requires code execution to be enabled"; sharing with colleagues and groups on Team and Enterprise (read 2026-11-05)
  4. anthropics/skills: skill-creator (read 2026-11-05)
  5. Agent Skills specification — name and description required; description up to 1,024 characters (read 2026-11-05)