# How to Create Claude Skills for Claude and Claude Code (2026)

> A Claude skill is a folder with a SKILL.md file: a short YAML header with a name and a description, followed by plain-language instructions. In Claude Code, save it as ~/.claude/skills/<skill-name>/SKILL.md for yourself or .claude/skills/<skill-name>/SKILL.md in a project, and run it with /skill-name or let Claude load it when a request matches. In the Claude app, turn on code execution, zip the folder and upload it under Customize > Skills.

Format: howto · Published: 2026-11-20
URL: https://cognovaai.com/insights/how-to-create-claude-skills

Skills let you teach Claude a repeatable task once, such as how your firm writes a client brief, and reuse it every time. "Claude code skills" is searched about 9,900 times a month in the US and "claude skills" about 40,500, according to Semrush, yet most top results are forum threads and third-party explainers rather than step-by-step instructions.

This guide follows Anthropic's Claude Code docs, the Agent Skills docs and Anthropic's help center as published on November 20, 2026. It is written for business owners and the staff who build internal tools. We built and validated the example skill below on our own Windows 11 PC.

## What you need

- Claude Code, or a Claude account on Free, Pro, Max, Ultra, Team or Enterprise with code execution turned on
- One repeatable task you can describe step by step, with an example of a good result
- A plain-text editor
- Optional: any reference documents or scripts the task needs

**Time needed:** 30–60 minutes for a first skill

## Step 1: Pick one repeatable task

Choose a task you or your staff do the same way every week, such as summarising client intake forms, drafting a proposal from notes, or checking a document against a checklist. Anthropic recommends one focused skill per workflow. Write down the steps, the format of a good result and the mistakes to avoid. Keep two or three real examples to test with later.

## Step 2: Create the skill folder and SKILL.md

Make a folder named after the skill, using lowercase letters, numbers and hyphens, for example `client-intake-summary`. Inside it, create a file called `SKILL.md`. The file must start with a YAML header between `---` lines containing `name` and `description`. Anthropic's Agent Skills docs limit `name` to 64 characters and say it cannot contain the words "anthropic" or "claude".

## Step 3: Write a description that says what and when

Claude uses the description to decide when to load the skill, so it matters more than anything else. Say what the skill does and when to use it, in the third person, for example "Summarizes a new client intake form into a one-page brief. Use when the user shares an intake form." The Agent Skills docs allow up to 1,024 characters, but Anthropic's help article on uploads gives 200, so stay under 200 if you plan to upload the skill.

## Step 4: Write the instructions

Below the header, write the steps in plain language, as you would brief a capable new employee. Anthropic's best-practice guide says Claude is already very capable, so include only what it would not know: your format, your rules and your examples. Keep SKILL.md under 500 lines. Put long reference material in separate files in the same folder, link to them from SKILL.md, and use forward slashes in paths, even on Windows.

## Step 5: Use the skill in Claude Code

Save the folder in `~/.claude/skills/` to use it in all your projects (on Windows this is `C:\Users\<you>\.claude\skills\`), or in `.claude/skills/` inside a project to share it with your team through Git. Type `/skills` to see what Claude Code has loaded, run the skill directly with `/client-intake-summary`, or ask for something that matches the description. You can also simply ask Claude "What skills are available?"

## Step 6: Validate the file

Run `claude plugin validate ~/.claude/skills` to catch SKILL.md files whose header does not parse. We ran this check on our example skill on November 20, 2026 and it printed "Validation passed". A header that is not on the very first line of the file is a common reason a skill is ignored, because Claude Code reads the header only when the opening `---` is the first line.

## Step 7: Upload the skill to the Claude app

On an individual plan, go to Settings > Capabilities and turn on "Code execution and file creation" (called "Create and edit files" in the newer layout). Zip the skill folder so the folder itself is at the root of the zip, for example `client-intake-summary.zip` containing `client-intake-summary/SKILL.md`. Then go to Customize > Skills, click **+**, choose **+ Create skill**, then **Upload a skill**, and switch the skill on.

## Step 8: Test it and review security

Try the skill with your saved examples and a few prompts that should and should not trigger it, then tighten the description and instructions. Anthropic advises using skills only from trusted sources, such as ones you wrote or ones from Anthropic, because a malicious skill can direct Claude to run code or call tools in ways that do not match its stated purpose. Never put passwords or API keys in a skill.

## An example skill you can copy

This is the skill we created and validated with `claude plugin validate` on November 20, 2026. Save it as `client-intake-summary/SKILL.md`:

```markdown
---
name: client-intake-summary
description: Summarizes a new client intake form into a one-page brief with contact details, the request, deadlines and open questions. Use when the user shares an intake form or asks for a client brief.
---

# Client intake summary

1. Read the intake form the user provides.
2. Write a one-page brief with these headings: Client, Request, Deadline, Budget, Open questions.
3. List anything missing from the form under Open questions. Do not guess missing values.
```

The description is under 200 characters, so the same folder can be zipped and uploaded to the Claude app.

## Optional settings in Claude Code

Claude Code accepts more header fields than the Claude app. Two are useful for business workflows:

- `disable-model-invocation: true` stops Claude from loading the skill on its own, so it runs only when someone types `/skill-name`. Use this for skills that send or change things.
- `allowed-tools` limits which tools the skill may use.

Skills uploaded to the Claude app accept only `name`, `description`, `license`, `compatibility`, `metadata` and `allowed-tools`, according to Anthropic's Claude Code docs.

## Tips from Anthropic's best-practice guide

- Write the description in the third person ("Processes Excel files and generates reports"), not "I can help you".
- Prefer clear names that describe the activity, such as `processing-invoices`, over vague ones like `helper`.
- Keep links to other files one level deep from SKILL.md.
- Test with the models your team actually uses, and compare results with and without the skill.

## What we verified

On a Windows 11 PC with Claude Code 2.1.312, we created the example skill above and ran `claude plugin validate` on its folder, which passed. We did not upload it to a Claude app account. The app upload steps come from Anthropic's help center as published on November 20, 2026.

## How Cognova AI can help

We build Claude skills, connectors and automations around how your firm already works, then train staff to use and maintain them. See [AI Consulting](/services/ai-consulting-dallas), [AI for CPA Firms](/services/ai-os-cpa-firms), or our guide to [Claude Cowork](/insights/how-to-use-claude-cowork).

## Frequently asked questions

### What is a Claude skill?

A skill is a folder containing a SKILL.md file with a name, a description and instructions, plus any reference files or scripts the task needs. Claude reads only the short name and description until a request matches, then loads the full instructions. This lets you give Claude many specialised abilities, such as your firm's report format, without filling every conversation with long instructions.

### Which Claude plans support skills?

Anthropic's help center lists skills on Free, Pro, Max, Ultra, Team and Enterprise plans, with code execution turned on. On Team and Enterprise plans an owner must enable code execution and skills under Organization settings. In Claude Code, skills work on any account that can run Claude Code. Skills you enable on claude.ai also sync into Claude Code version 2.1.273 or later when you sign in with the same account.

### What is the difference between a skill and a custom instruction?

Custom instructions apply to every conversation. A skill loads only when a request matches its description or when you run it by name, and it can include extra files and scripts. Use custom instructions for general preferences, such as tone and date format, and skills for specific repeatable tasks, such as turning intake forms into client briefs or checking contracts against a checklist.

## Sources

- [Anthropic: Extend Claude with skills (Claude Code docs)](https://code.claude.com/docs/en/skills)
- [Anthropic: Agent Skills overview](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview)
- [Anthropic: Skill authoring best practices](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices)
- [Anthropic Help Center: Using skills in Claude](https://support.claude.com/en/articles/12512180-using-skills-in-claude)
- [Anthropic Help Center: How to create custom skills](https://support.claude.com/en/articles/12512198-creating-custom-skills)
- [Anthropic example skills (GitHub)](https://github.com/anthropics/skills)
- [Semrush Keyword Overview: claude code skills (US database, November 2026)](https://www.semrush.com/analytics/keywordoverview/?q=claude+code+skills&db=us)
