Reusable prompts in Cursor: skills, rules and commands

Which goes where

From Cursor's help and docs, read 5 November 2026.

A worked example

Illustration: a made-up business. Ken, a developer on a 5-person team, typed the same prompt every time he wanted tests: use Vitest, one file per module, cover the error paths, no snapshot tests. He turns it into a skill:

• Folder: .cursor/skills/write-tests/, committed to the repository.
• Description: "Writes Vitest tests for a module in our style. Use when someone asks for tests, test coverage or a test file."
• Steps: his four rules, plus the correction he kept giving: "mock the network, never the database".
• One example: a short module and the test file he'd want.
• Checks: every exported function has a test; error paths are covered; no snapshot tests; it runs with npx vitest run.

"Use the team logger" stays a rule, set to Always Apply, because it's one line that applies everywhere. He tests the skill on three modules, including one with no error handling, and fixes the steps where the draft went wrong.

Set yours up

1. List the prompts you retype and the one-line rules you repeat.
2. Make each prompt a skill folder with a "use when" description.
3. Make each one-line rule a rule, with the type that fits.
4. Run /migrate-to-skills if you have old slash commands.
5. Commit the folders so teammates get the same skills.

Next step: the same skills for the whole team

Skills in the repository reach everyone who works in it, whichever tool they use. To get the whole team working the same way, see getting your team to use AI the same way, and for keeping one set of skills in step across Claude Code, Codex, Cursor and OpenCode, syncing skills across tools.

Questions

How do I save a reusable prompt in Cursor?

Save it as a skill: a folder such as .cursor/skills/write-tests/ with a SKILL.md inside. Run it by typing /write-tests in chat, or attach it as context with @write-tests. Short guidance that should apply all the time belongs in a rule instead.

What is the difference between Cursor rules and skills?

Cursor's help puts it simply: use a rule when a short instruction is enough, and a skill when the agent needs a detailed, repeatable process. Rules are included in every conversation, or matching ones. Skills are applied when the agent finds them relevant, or when you type /name; set disable-model-invocation: true and a skill runs only when you call it (Cursor Docs: Skills).

What happens to my old Cursor slash commands?

From Cursor 2.4, typing /migrate-to-skills in Agent chat converts your user-level and workspace-level slash commands into skills, keeping them manual, along with rules that aren't always-on or tied to file patterns.