Creating Claude Code Skills FAQs

Question 1

How does this skill improve token efficiency?

Accepted Answer

It teaches a 'progressive loading' architecture. By using companion files (like REFERENCE.md or TROUBLESHOOTING.md), Claude only loads the content it needs for a specific query, preventing large instruction sets from consuming unnecessary context window space.

Question 2

What are the core components of a Claude Code skill?

Accepted Answer

A skill is filesystem-based, requiring a directory with a 'SKILL.md' file. This file must contain YAML frontmatter with a 'name' and 'description' to help Claude automatically discover and trigger the skill when relevant to your tasks.

Question 3

What does the 'Creating Claude Code Skills' skill do?

Accepted Answer

This is a meta-skill designed to guide developers through the process of building their own Claude Code skills. It provides standards for skill anatomy, naming conventions, organization patterns, and best practices for extending Claude's native capabilities.

Question 4

Where should I store my custom Claude Code skills?

Accepted Answer

Storage depends on portability: use '~/.claude/skills/' for personal/portable skills, '~/.claude/skills.private/' for company-specific workflows not in version control, and 'project/.claude/skills/' for team-shared, project-specific procedures.

Question 5

How do I know if my skill is too complex and needs refactoring?

Accepted Answer

This guide identifies 'red flags' for refactoring, such as when a skill exceeds 800 lines, covers more than three distinct tools, or contains troubleshooting sections that are significantly longer than the core instructions.

Creating Claude Code Skills

Creating Claude Code Skills

주요 기능

사용 사례

주요 기능

사용 사례