AI With Amirthan

← All guides

Claude Code · July 11, 2026 · 3 min read

Turn any workflow you repeat into a Claude Code skill

If you explain a workflow to Claude twice, make it a skill. What a SKILL.md is, why the description line matters, and a prompt that writes your first one.

The problem

Some prompts you type once and forget. Others you find yourself re-explaining every week: the checklist for publishing a post, the steps to cut a release, the audit you run before a deploy. Claude does the work fine, but the instructions live in your head, and every retelling drifts a little. The third version of the workflow is never quite the first. In Claude Code, that repeated explanation is exactly what a skill is for.

What is a Claude Code skill?

A skill is a folder with a SKILL.md file inside: plain markdown instructions that Claude Code loads on demand. Put it at .claude/skills/<name>/SKILL.md in a repo (or ~/.claude/skills/ to use it across all your projects) and it becomes a slash command, /name, that Claude can also pull in on its own when your task matches.

The underrated part is the frontmatter description. Between uses, the description is the only piece of the skill sitting in Claude’s context; the full instructions load only when the skill is invoked. So skills cost almost nothing while idle, and the description is what decides whether one ever fires. Write it to say what the skill does and when to use it, in the words you would actually type.

I run this site through skills checked into the repo. new-guide holds my whole guide-writing contract: frontmatter rules, voice, word limits, the SEO checklist. seo-check audits every built page for one h1, title and description lengths, and internal links, then fixes the small stuff itself. The guide you are reading was written under that contract. The same pattern eventually scaled into rendering Instagram carousels from one command.

Anything you explain to Claude twice should become a skill.

Turn your last repeated workflow into a skill

The trigger is simple: the second time you explain a workflow, stop and capture it. The handoff prompt for long sessions is a perfect first candidate. Paste this into Claude Code inside the repo where the workflow lives:

Turn the workflow I just walked you through into a Claude Code skill.
1. Create .claude/skills/<short-name>/SKILL.md.
2. Frontmatter: name, plus a one-line description that says what the
   skill does AND when to use it, phrased in words I would say.
3. Body: the exact steps, commands, file paths, and checks, written
   so a fresh session can follow it without asking me anything.
4. Ask me about any step you are unsure of. Show me the file before
   saving it.

Do this now

  1. Name the last workflow you explained to Claude for the second time.
  2. Paste the prompt above in that repo and answer its questions.
  3. Run your new slash command once, fix anything that drifted, and commit the folder so the skill travels with the repo.

Claude Code

Turn this into a Claude skill

Paste this into Claude Code. It reads this guide and builds a skill shaped to how you work. 

Read the guide at https://aiwithamirthan.com/guides/turn-any-workflow-into-a-claude-code-skill. Then build me a Claude Code skill from it, personalized to how I actually work.

1. Fetch and read the full guide at the URL above.
2. Look at my setup so the skill fits me, not a generic template: skim my repos, my CLAUDE.md / AGENTS.md files, my existing skills, and my stack.
3. Create a new skill (a SKILL.md with a clear name, description, and steps) that lets me do what this guide teaches, adapted to my tools and goals.
4. Tell me where you saved it and how to invoke it.

If anything about my setup is unclear, ask me one or two questions first.

Want to build this yourself?

Join Claude Code Academy: courses, build alongs, and builders turning AI into real systems. And I post a new build on Instagram nearly every day.

Join Claude Code Academy → Follow @aiwithamirthan