Claude skills examples: real SKILL.md files, explained
Claude skills examples you can open and read: the anatomy of a real SKILL.md, Anthropic’s example skills by task, and visual skills shown with their output.
- For
- Anyone who wants to see what a skill looks like before using or writing one
- Written for
- Claude CodeCodexCursorGitHub CopilotGemini CLIOpenCodeAny agent that reads SKILL.md folders
- Time
- 7 min read · updated Nov 10, 2026
The short answer
Good Claude skills examples are easy to find once you know the shape. A skill is a folder with a SKILL.md file: a short header with a name and a description that says what it does and when to use it, then step-by-step instructions, plus any scripts and templates those steps call. This page takes one real skill apart line by line, then lists Anthropic’s public example skills and Lookbk’s free visual skills, whose pages show what they rendered.
Agent skills examples look the same whichever agent reads them, because the format is shared: the Agent Skills specification defines a skill as a folder containing a SKILL.md file, with optional scripts/, references/ and assets/ folders beside it. Claude Code, Codex, Cursor and other agents load that format. So an example written for Claude is also an example for the others.
Anatomy of a Claude skill: one real example
The worked example is Editorial OG Card, a free Lookbk skill that makes 1200×630 link-preview images for blog posts. Its folder, trimmed to the parts that matter here:
1. The header: name and description
Agents read only this header until a request matches, so it carries the most weight. The name is lowercase with hyphens and matches the folder; the spec allows up to 64 characters. The description (up to 1,024 characters) does two jobs: the first sentence says what the skill makes, and the last says when to use it, using the words people actually type: “OG images, social cards, link previews or Twitter/X cards”. Claude Code’s docs say Claude uses the description to decide when to apply a skill. Write a description that makes a skill trigger goes deeper.
2. The instructions: what you need, then numbered steps
The body opens with a plain list of what the output looks like and a What you need section (Node.js 18 or newer, and Chrome, Chromium or Edge). Then come six numbered steps, each one an action:
Each step gives the exact command and what to check. Step 3, for example:
Followed by what it does: “It checks every card first and lists every problem at once, then renders the cards four at a time … and checks every PNG is 1200×630.” The agent doesn’t have to guess whether it worked.
3. Scripts and templates: the parts the model doesn’t improvise
The steps call scripts rather than asking the model to write code each time. from-posts.mjs reads your posts’ front matter into cards.json, render.mjs checks every field against its limit and takes the screenshots, and template/ holds the card design. Because the design is a file, the tenth card matches the first. The spec calls scripts/ the place for executable code, and asks that scripts document their dependencies and give helpful error messages.
4. Troubleshooting and a file list
The skill ends with fixes keyed to what you will actually see, such as render: fix cards.json first or a headline drawn in a plain serif because the font stylesheet didn’t load, and a Files list saying what each script does. That is what lets an agent recover from an error without asking you.
Example skills from Anthropic
Anthropic’s public anthropics/skills repository is the other place to read real skills. Its README groups them as creative and design, development and technical, enterprise and communication, and document skills, and says they are for demonstration and education. Here are the ones with a clear single job, grouped by task, with their descriptions summarized:
Read the folders as well as the descriptions. pdf carries forms.md, reference.md and a scripts/ folder, and skill-creator has agents/, references/, scripts/ and an eval viewer: the long material sits in files the agent opens only when a step needs it. To try them in Claude Code:
Visual skill examples you can see first
Skills that make something visual are easiest to judge by their output. Every Lookbk skill page shows what the skill rendered from its example briefs. These are free:
What makes a good example skill
Whether you are choosing a skill or writing one, the strong examples above share the same habits.
A skill that works
- A description that says what it makes and when to use it, in the words people type
- A What you need list before the first step
- Numbered steps, each with the exact command to run
- Checks done by a script, with every problem listed at once
- Fixes keyed to real error messages
- A Files list, and long reference material in separate files
A skill that drifts
- A one-line description such as “Helps with images”
- Requirements you only discover when a step fails
- Advice in paragraphs with no order to follow
- “Make sure it looks good” in place of a check
- No word on what to do when something breaks
- One very long
SKILL.mdwith everything in it
Both the Agent Skills spec and Claude Code’s docs say to keep SKILL.md under 500 lines and move detailed reference material to separate files. How to create a Claude skill builds one from scratch.
Try one
Free Lookbk skills install with no account. Run the installer, start a new session, and ask for the job in plain words, such as “Make Editorial OG cards for every post in /content/blog”.
Then open the installed SKILL.md and read it next to this page. For other places to find skills, see Claude skills marketplace.
Common questions
- What is an example of a Claude skill?
- Anthropic’s
pdfskill is one: a folder whoseSKILL.mdtells Claude how to extract, merge, split and fill PDFs, with scripts for the heavy work. Editorial OG Card is another, which turns blog posts into 1200×630 link-preview images. - What goes in a SKILL.md file?
- A YAML header with at least a
nameand adescription, then Markdown instructions. The Agent Skills spec suggests step-by-step instructions, examples of inputs and outputs, and common edge cases. - Does a skill need scripts?
- No. A skill can be instructions only. Scripts help when a step must come out the same every time, such as checking lengths or rendering an image.
- Where can I find more Claude skills examples?
- Anthropic’s
anthropics/skillsrepository on GitHub, the plugin marketplaces in Claude Code, open directories such as skills.sh, and catalogs such as Lookbk. Claude skills marketplace compares them. - Are agent skills examples different from Claude skills examples?
- Not in format. Agent Skills is a shared format, so a
SKILL.mdfolder written for Claude can be loaded by Codex, Cursor and other agents that read it, from their own skills folders.
Sources
What this guide says about other companies’ tools comes from their own documentation, checked on Nov 10, 2026.
Did this guide help?
Read next
All guidesWhat are agent skills?
A plain-English guide to SKILL.md folders: what is inside one, how an agent decides to use it, and when a skill beats a long prompt.
How to create a Claude skill, step by step
Create a Claude skill from an empty folder: write SKILL.md, add scripts, check it triggers, then share it and make it work in Codex and Cursor too.
Write a description that makes a skill trigger
How an agent picks a skill, what to put in its description, and how to test that it triggers when it should and stays quiet when it shouldn’t.
Claude skills marketplace: where to find skills in 2026
There’s no one Claude skills marketplace. Where to find skills: Anthropic’s repo and plugin marketplaces, open directories and curated catalogs, compared.
See a skill’s result before you install it.
Every skill page shows what it makes, what it needs and which agents can run it.