Cursor skills and custom slash commands: .cursor/skills SKILL.md, .cursor/commands and /migrate-to-skills

• Project: .cursor/skills/ and .agents/skills/. Each skill is a folder with a SKILL.md file inside.
• User (all your projects on this machine): ~/.cursor/skills/ and ~/.agents/skills/.
• For compatibility Cursor also loads .claude/skills/, .codex/skills/, ~/.claude/skills/ and ~/.codex/skills/, so skills written for Claude Code or Codex work without copying.
• Monorepos: a .cursor/skills/ or .agents/skills/ folder inside a subdirectory such as apps/web/ is found too, and its skills apply only to files under that directory.
• You can group skills in category folders, such as .cursor/skills/shipping/deploy-staging/SKILL.md. The skill name comes from the folder that holds SKILL.md.

Cloud Agents only get your ~/.cursor/skills/ folder if you turn on Sync Skills for Cloud Agents in Settings, under Agents. ~/.agents/skills/ and unsynced local skills are not copied to Cloud Agents, remote SSH sessions or self-hosted workers, so put skills a team needs in the repo.

• name (required): lowercase letters, numbers and hyphens. It must match the parent folder name.
• description (required): what the skill does and when to use it. The agent reads this to decide whether the skill is relevant.
• paths (optional): glob patterns that limit the skill to matching files, as a list or a comma-separated string. Without it the skill is available no matter which files are open. The older globs field still works as a fallback.
• disable-model-invocation (optional): when true, the skill is used only when you call it with /skill-name.
• icon and color (optional): the badge shown when you use the skill as a Custom Mode.
• metadata (optional): any extra key-value pairs.

---
name: react-component-patterns
description: Conventions for writing React components in this codebase.
paths:
  - "**/*.tsx"
  - "packages/ui/**/*.ts"
---

# React component patterns

Use function components, colocate tests, and export one component per file.

A command is a skill the agent should not pick on its own. Set disable-model-invocation: true, then type / in Agent chat and choose it. A skill you call by hand applies to that one message.

---
name: review
description: Review the current changes for bugs and missing tests.
disable-model-invocation: true
---

Review the current changes for bugs and missing tests.
List problems in order of severity.

• Type / and the skill name, for example /review, to run it.
• Type @ and select the skill to attach it as context instead.
• To keep a skill on for the whole session, use it as a Custom Mode with Option+Enter on Mac or Alt+Enter on Windows.
• /create-skill runs a built-in guide that writes a new skill for you.

Cursor 1.6 added custom slash commands as Markdown files in .cursor/commands/[command].md, run by typing / in the Agent input. Current Cursor docs describe commands only as something to convert into skills.

• Type /migrate-to-skills in Agent chat (Cursor 2.4 or later). This built-in skill finds eligible rules and commands and converts them.
• Converted: user-level and workspace-level slash commands, which keep their explicit invocation through disable-model-invocation: true, and dynamic rules (Apply Intelligently rules with alwaysApply false or unset and no globs).
• Not converted: rules with alwaysApply: true, rules with specific globs, and user rules, which are not stored as files.
• Review the generated skills in .cursor/skills/ before you delete the originals.

• scripts/: code the agent can run, such as scripts/deploy.sh <environment>.
• references/: longer documentation the agent loads only when needed.
• assets/: templates, images or data files.
• Keep SKILL.md short and move detail into these folders, because Cursor loads them progressively.

• Rules: short guidelines and constraints that Cursor adds to every matching conversation.
• Skills: longer, repeatable multi-step procedures that the agent loads when relevant or that you call with /skill-name.
• Commands: in current Cursor, a skill with disable-model-invocation: true.

To share a personal skill with your team, open Customize, then Skills, open the skill and choose Publish. It goes to your team's Default marketplace, where teammates can install it.

The Cursor CLI (agent) has its own built-in commands. Its reference page does not describe custom commands. Some of the most used ones:

• /model [filter]: pick a model.
• /run-everything [on|off|status]: toggle Run Everything. /auto-run is an alias.
• /plan, /ask and /debug: switch to Plan, Ask (read-only) or Debug mode.
• /mcp [list|list-tools]: manage MCP servers. /plugin manages plugins and marketplaces.
• /sandbox: configure sandbox mode and network access.
• /summarize (alias /compress), /resume, /fork and /rewind: manage the conversation.
• /help [command]: help for any command.

• Compare: custom slash commands in every AI coding agent
• Compare: Agent Skills (SKILL.md) in every AI coding agent
• Cursor rules setup: project, user and Team Rules
• Cursor subagents: .cursor/agents
• Cursor allow all commands: Run Everything and allowlist
• Cursor CLI: headless mode and permissions