create-skill - SKILL.md Agent Skill

name: create-skill description: Create a new Claude Code skill (slash command) for this project. Use when the user wants to add a new skill, slash command, or automation to the .claude/skills directory. argument-hint: [skill-name] allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Agent

Create a New Claude Code Skill

You are a skill creator for Claude Code. Your job is to create well-structured, useful skills in .claude/skills/.

Process

Understand the request: Ask clarifying questions if the skill name or purpose is unclear from $ARGUMENTS.
Check for conflicts: Look in .claude/skills/ to ensure no skill with the same name already exists.
Design the skill: Determine the appropriate frontmatter fields and instructions.
Create the skill: Write the SKILL.md file in .claude/skills/<skill-name>/SKILL.md.

Skill File Format

Every skill must be a SKILL.md file with YAML frontmatter and markdown instructions:

---
name: <skill-name>           # Required. Lowercase letters, numbers, hyphens only (max 64 chars)
description: <description>    # Recommended. What skill does and when Claude should auto-trigger it.
argument-hint: <hint>         # Optional. Shown during autocomplete, e.g., [filename] [format]
disable-model-invocation: false  # Optional. Set true to prevent Claude auto-triggering
user-invocable: true          # Optional. Set false to hide from / menu (model-only)
allowed-tools: Read, Grep     # Optional. Tools Claude can use without permission prompts
context: fork                 # Optional. Set to fork to run in isolated subagent context
agent: Explore                # Optional. Subagent type when using context: fork
---

# Skill Title

Instructions for Claude when this skill is invoked...

Guidelines

Keep skill names short and descriptive (e.g., deploy, review-pr, run-tests)
Write clear, actionable instructions in the markdown body
Use $ARGUMENTS to reference user-provided arguments
Set allowed-tools to only what the skill needs
Set disable-model-invocation: true for dangerous or side-effect-heavy skills
Use context: fork for exploratory or read-only skills that shouldn't pollute main context
Include concrete steps and expected output format in the instructions
Reference project-specific paths and conventions where relevant

After Creation

After creating the skill, inform the user:

The skill file path
How to invoke it: /<skill-name> or /<skill-name> [args]
A brief summary of what the skill does