create-skill

/create-skill - Skill Generator

$ARGUMENTS

---

Purpose

Generate a new Claude Code skill (slash command) from a natural language description. Creates properly structured SKILL.md with frontmatter, behavior rules, and output format.

---

Behavior

When /create-skill is triggered:

Step 1: Parse Input

Extract from user's description:

• Skill name: lowercase, hyphenated (e.g., api-test, changelog)
• Purpose: what the skill does
• Trigger terms: when should it activate

If input is vague, ask:

1. What should this skill do?
2. Does it modify code or only analyze/report?
3. What sub-commands should it have?

Step 2: Determine Skill Type

Type	Characteristics	disable-model-invocation
Analysis	Read-only, generates reports	true
Generator	Creates new files/code	true
Interactive	Asks questions, explores options	false
Automation	Runs scripts, orchestrates agents	true

Step 3: Generate SKILL.md

Create .claude/skills/{name}/SKILL.md with this structure:

---
name: {skill-name}
description: >
  {One-line purpose}.
  Use when: {comma-separated trigger terms}
argument-hint: "{expected arguments}"
user-invocable: true
disable-model-invocation: {true|false}
---

# /{skill-name} - {Title}

$ARGUMENTS

---

## Purpose

{What this skill does and why it exists.}

---

## Sub-commands

{If applicable}

---

## Behavior

When `/{skill-name}` is triggered:

### 1. {First Step}
{Description}

### 2. {Second Step}
{Description}

### 3. {Third Step}
{Description}

---

## Output Format

{Expected output template}

---

## Examples

{Usage examples}

---

## Key Principles

- {Principle 1}
- {Principle 2}
- {Principle 3}

Step 4: Validate

Check generated skill:

• Name is lowercase, hyphenated, max 30 chars
• Description fits in 250 chars (budget limit)
• Has clear trigger terms in description
• $ARGUMENTS placeholder is present
• No unsupported frontmatter fields
• Output format is defined