Anthropic의 공식 문서에서 가장 좋은 관행을 가진 Claude Code 기술을 창조합니다.

왜 기술?

Claude는 한 번에 모든 것을 붙들 수 없습니다. 스쿼시!

Skills는 정보를 점차적으로 풀어 놓은 대리인의 문제를 해결합니다: 1.Startup: 이름 + 설명 로드(~100개의 토큰) 2. ** 트리거 될 때 : 적재되는 가득 차있는 지시 3. ** 필요 **: 작업이 필요할 때만 로드 된 참조

이 문제 해결에 대한 풍부한 지식을 만들기 동안 컨텍스트 린 유지.

Skills vs slash 명령:

• Skills are model-invoked (Claude apply them when related based on frontmatter and then related link in SKILL.md).
• Slash 명령은 user-invoked (/command)입니다.

빠른 시작

1. Claude Code에 있는 이 폴더를 엽니다
2. 새로운 기술을 만들려면
3. 몇 가지 질문에 대답 - 참조 또는 온라인 검색에 클레ude에 대한 포인트
4. output_skills/[category]/[skill-name]/에서 기술 찾기

당신의 능력을 사용

이 repo는 ./skills] 글로벌 설치에 대한 헬퍼 스크립트가 포함되어 있습니다.

Global (모든 프로젝트) - symlink에 스크립트를 사용 ~/.claude/skills/:

output_skills/] symlink 이므로 즉시 수정합니다.

Skill Factory

Creates Claude Code skills with built-in best practices from Anthropic’s official documentation.

Why Skills?

Claude can’t hold everything in mind at once. Squirrel!

Skills solve the problem of distracted agent by releasing information gradually:

1. Startup: Only name + description loaded (~100 tokens per skill)
2. When triggered: Full instructions loaded
3. As needed: References loaded only when the task requires them

This keeps context lean while making rich knowledge available on demand.

Skills vs slash commands:

• Skills are model-invoked (Claude applies them when relevant based on the frontmatter and then relevant links in SKILL.md).
• Slash commands are user-invoked (/command).

Quick Start

1. Open this folder in Claude Code
2. Ask it to create a new skill
3. Answer a few questions — point to references or ask Claude to search online
4. Find your skill in output_skills/[category]/[skill-name]/

Using Your Skill

This repo includes a ./skills helper script for global installation.

Global (all projects) — use the script to symlink to ~/.claude/skills/:

./skills toggle    # interactive picker
./skills status    # check what's installed

Edits to output_skills/ apply immediately since it’s a symlink.

Local (single project) — copy the skill folder to your project’s .claude/skills/ or create a symlink in your project yourself.

STARTER_CHARACTER

Each skill defines a STARTER_CHARACTER = [emoji] at the top. This is a visual indicator that Claude has loaded the skill and is following its instructions. For example, the nullables skill uses ⭕️ and TDD uses 🔴/🌱/🌀.

To activate this, add the following to your global ~/.claude/CLAUDE.md:

Always start replies with STARTER_CHARACTER + space (default: 🍀). Stack emojis when requested, don't replace.

Pick any default emoji you like — it confirms Claude is reading your ground rules. When a skill activates, its emoji stacks on top: 🍀 ⭕️ means both your global rules and the nullables skill are active.

Without this line in your global CLAUDE.md, the STARTER_CHARACTER lines in skills might have unpredictable behavior.

Updating Best Practices

./update-docs

Pulls latest skill patterns from Anthropic.

Structure

docs/           — Skill creation knowledge and patterns
output_skills/  — Generated skills organized by category