What is a Claude skill?

A Claude skill is a set of instructions you save so Claude can use the same method again. A prompt might ask for this week’s release notes. A skill tells Claude how your team decides what belongs in them, every week.

What goes in a skill?

Open a skill folder and look for SKILL.md. At the top are a name and a description of when to use it. The rest explains how to do the work. The folder can also hold examples, templates, or scripts; a written checklist alone is enough.

In Claude Code, you can invoke a skill by typing a slash followed by its name. Claude can also select a relevant skill from its description. Anthropic’s Claude Code guide explains where to put the folder and how to control when it runs.

Try the release-note test.

Take two engineering notes: “Fixed blank CSV downloads when a date filter is active” and “Renamed the export helper.” Which one belongs in a customer update?

A useful rule for a release-note skill is: write what changed for the reader, keep the condition that caused the problem, and leave out internal renames.

That gives you a line you can check: “CSV downloads now include your results when a date filter is active.” It doesn’t promise that every export bug is fixed. The same rule will still be useful when next week’s notes are different.

This is a worked example, not a recorded Claude run or a skill for sale. For another example, compare two CV bullets written from the same notes.

Will a Claude skill work in OpenClaw?

Sometimes. An OpenClaw skill also uses SKILL.md. A checklist you apply to pasted notes is easier to move between tools than a skill that calls a particular script. The same filename doesn’t guarantee the same result.

For example, if the instructions tell Claude to call a tool by name, check that your OpenClaw setup actually has it. Our guide to choosing an OpenClaw skill on ClawHub walks through checking a release and trying it on meeting notes.