Commands

Commands are skills invoked explicitly by the user via /name, never triggered automatically by the model. Use them for workflows that should only run when you choose — deployments, releases, or any operation where accidental auto-invocation would be disruptive.

Regular skills description field is used by agents to trigger the skill automatically. Commands suppress that automatic invocation:

	Auto-invoked by model	User-invoked via `/name`
Skill	✅ when description matches	✅
Command	❌	✅

Two mechanisms

1. `commands/` folder (legacy, wider support)

Place a markdown file in .claude/commands/ (project) or ~/.claude/commands/ (user):

.claude/commands/
  deploy.md       → /deploy
  release.md      → /release

The file content is the skill body. Frontmatter (description, allowed-tools) is supported. The model never auto-invokes these — they are user-only by design.

allowed-tools pre-approves the tools it lists so the command’s turn runs without permission prompts. It does not restrict the command to those tools: everything else stays callable, subject to your normal permission settings. To take a tool away, use disallowed-tools.

Claude Code marks commands/ as deprecated in favor of SKILL.md files, but it still works and is the more portable choice today (see harness compatibility below).

2. `disable-model-invocation: true` (modern, not universal)

Add the frontmatter field to any SKILL.md:

---
name: deploy
description: Deploy the application to production
disable-model-invocation: true
---

This prevents automatic loading while keeping the skill accessible via /deploy. It is the recommended approach in Claude Code’s own documentation, but not all harnesses support it.

Harness compatibility

Standalone skills

Harness	`disable-model-invocation` in `SKILL.md`	`commands/` folder
Claude Code	✅	✅ (deprecated, still works)
Cursor	⚠️ bug: hides plugin-delivered skills from `/` menu (Mar 2026)	—
Windsurf	✅	—
GitHub Copilot CLI	❌ (skills)	✅ (reads `.claude/commands/`)
Codex CLI	❌ (uses `agents/openai.yaml` instead)	—
Gemini CLI	❌	—

Plugin commands

Inside a universal plugin, a commands/ subfolder creates user-invocable slash commands across harnesses:

Component	Claude Code	Cursor	Codex
`skills/`	✅ native	✅ native	✅ native
`commands/`	✅ native	✅ native	silently ignored
`agents/`	✅ native	✅ native	silently ignored
`rules/`	silently ignored	✅ native	silently ignored

Because Cursor supports commands/ natively in plugins, and the disable-model-invocation: true frontmatter field has a known bug in Cursor (plugin-delivered skills with this flag are hidden from the / menu), the commands/ folder is the safer choice for cross-harness plugin commands targeting both Claude Code and Cursor.

Recommendation: Use commands/ inside your plugin for user-only workflows. When the Cursor bug is resolved and disable-model-invocation: true has stable cross-harness support, you can migrate commands to skills/ with the frontmatter flag.

Agent Configuration — the full picture of skills, commands, and rules
Universal Plugin governance — plugin component table and distribution
Claude Code skills docs — disable-model-invocation reference