command-writing

Command Writing Skill

A guide for creating effective Cursor commands - simple, reusable workflows triggered with the / prefix.

When using this skill, begin by stating: "I'm using the command-writing skill to create effective Cursor commands."

What is a Command?

A command is a simple Markdown file that defines a reusable workflow. Commands are:

• Explicit - Only triggered when user types /command-name
• Simple - Plain Markdown files without YAML frontmatter
• Focused - Single, specific workflow or task
• Parameterizable - Accept additional context after the command name
• Portable - Can be project-level, user-level, or team-level

Commands vs Skills

Understanding the distinction is critical:

Aspect	Commands	Skills
Invocation	Explicit via /command-name	Automatic (agent decides) or explicit
Format	Plain Markdown	Markdown with YAML frontmatter
Structure	Single .md file	Directory with SKILL.md + optional scripts/references
Complexity	Simple workflows	Complex, multi-faceted guidance
Parameters	Text after command name	Context-aware loading
Use Case	Specific, repeatable tasks	Domain knowledge and patterns

When to create a command:

• Single, specific workflow (e.g., "create PR", "run tests")
• User wants explicit control over invocation
• Simple enough to fit in one Markdown file
• No need for scripts, references, or progressive disclosure

When to create a skill:

• Complex domain knowledge (e.g., "story writing", "PR review")
• Agent should automatically apply based on context
• Needs scripts, references, or multiple variants
• Requires progressive disclosure for efficiency

Commands That Invoke Skills

Key principle: Commands define workflow structure, skills contain domain logic.

Critical: Avoid tight coupling to skill internals.

❌ Don't reference: skill sections, reference files, internal structure ✅ Do reference: capabilities, outcomes, what to accomplish