メニュー
Creates Claude Code skills or agents from a description. Use when the user wants to create, generate, build, or scaffold a new skill or agent.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/gravity9-tech/claude_code_workshop --skill craftコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Sets up Jira MCP server integration for Claude Code. Use when setting up Jira, connecting to Atlassian, configuring MCP for ticket management, or adding Jira capabilities.
Creates well-structured Claude Code slash commands from descriptions. Use when the user wants to create a new command, build an orchestrator, or make a workflow command that chains skills or agents.
Creates Claude Code hooks from descriptions. Use when the user wants to create a hook, add automation, set up pre/post tool actions, auto-format code, block commands, or add notifications.
SKILL.md を表示中
SOC 職業分類に基づく
| name | craft |
| description | Creates Claude Code skills or agents from a description. Use when the user wants to create, generate, build, or scaffold a new skill or agent. |
| argument-hint | <skill|agent> <description> |
Create a well-structured Claude Code skill or agent from a description.
$0 — artifact type: skill or agentIf $0 is not skill or agent, infer the type from context or ask the user.
Extract from $ARGUMENTS:
Use WebSearch to find domain best practices relevant to the artifact's purpose. This informs the instructions and tool selection.
Do NOT blanket-allow all tools. Select tools based on what the artifact actually does:
| Artifact needs to... | Grant these tools |
|---|---|
| Read/search code | Read, Grep, Glob |
| Modify files | Read, Edit, Write |
| Run shell commands | Bash (scope with patterns like Bash(git:*)) |
| Research online | WebSearch, WebFetch |
| Spawn subagents | Task |
| Nothing external | Omit allowed-tools entirely |
Principle: start with zero tools and add only what's justified by the instructions. If a tool isn't referenced in any instruction step, don't include it.
The description field is the most important metadata — it controls when Claude auto-loads the skill/agent.
Rules:
Good examples:
Reviews code for security vulnerabilities. Use when auditing code, checking for OWASP issues, or reviewing auth.
Sets up Playwright MCP server. Use when setting up browser automation, web testing, or scraping.
Bad examples:
A helpful skill for doing things with code. # too vague, no triggers
Security stuff. # no "when to use"
Read the appropriate reference file for the artifact type:
${CLAUDE_SKILL_DIR}/refs/skill.md for metadata fields, template, and validation checklist${CLAUDE_SKILL_DIR}/refs/agent.md for metadata fields, template, and validation checklistFollow the template structure and validate against the checklist in that file.
.claude/skills/<name>/SKILL.md or .claude/agents/<name>.md)/name for skills, @name for agents