home / skills / shunsukehayashi / miyabi / skill-creator

skill-creator skill

This skill helps you create well-structured Claude Skills for Miyabi projects by guiding setup, naming, and packaging workflows.

npx playbooks add skill shunsukehayashi/miyabi --skill skill-creator

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

3.5 KB

---
name: skill-creator
description: Create new Claude Skills for Miyabi project. Use when building custom skills, packaging workflows, or extending Claude's capabilities. Triggers on "create skill", "make skill", "new skill", "スキル作成", "スキルを作る".
allowed-tools: Bash, Read, Write, Grep, Glob
---

# Skill Creator

**Version**: 1.0.0
**Purpose**: Create properly structured Claude Skills

---

## Triggers

| Trigger | Examples |
|---------|----------|
| Create | "create a skill", "スキル作成", "make new skill" |
| Package | "package this workflow", "ワークフロー化" |
| Automate | "automate this process", "自動化" |

---

## Skill Structure

```
.claude/skills/[skill-name]/
├── SKILL.md              # Required: Main instructions
├── resources/            # Optional: Reference docs
│   └── examples.md
└── scripts/              # Optional: Helper scripts
    └── helper.sh
```

---

## Creation Process

### Step 1: Gather Requirements

Ask:
1. **Purpose**: What does this skill do?
2. **Triggers**: When should it activate?
3. **Input**: What information is needed?
4. **Output**: What should be produced?
5. **Tools**: Which tools are needed?

### Step 2: Create Directory

```bash
mkdir -p .claude/skills/[skill-name]
```

### Step 3: Write SKILL.md

```markdown
---
name: [skill-name]
description: [What it does]. Use when [trigger conditions].
allowed-tools: Bash, Read, Write
---

# [Skill Title]

**Version**: 1.0.0
**Purpose**: [Brief purpose]

---

## Triggers

| Trigger | Examples |
|---------|----------|
| [Category] | "[example1]", "[example2]" |

---

## Workflow

### Step 1: [Title]

[Instructions]

---

## Checklist

- [ ] [Item 1]
- [ ] [Item 2]
```

---

## Naming Rules

### Required

| Rule | Example |
|------|---------|
| Kebab-case | `my-skill` |
| Lowercase | `code-reviewer` |
| Descriptive | `test-generator` |

### Forbidden

| Word | Reason |
|------|--------|
| `claude` | Trademark |
| `anthropic` | Company name |
| `mcp` | Protocol name |

```
✅ GOOD: code-connector, task-runner
❌ BAD: claude-helper, mcp-server
```

---

## Description Format

```
[What it does]. Use when [conditions].
```

Examples:
```
✅ GOOD: "Review code for bugs and security issues.
         Use when checking PRs or analyzing code quality."

❌ BAD: "A skill for code"
❌ BAD: "Helps with stuff"
```

---

## Template

```markdown
---
name: [kebab-case-name]
description: [Action verb] [object]. Use when [trigger condition 1], [trigger condition 2].
allowed-tools: Bash, Read, Write, Grep, Glob
---

# [Title with Emoji]

**Version**: 1.0.0
**Purpose**: [One-line purpose]

---

## Triggers

| Trigger | Examples |
|---------|----------|
| [Category 1] | "[EN example]", "[JP example]" |
| [Category 2] | "[EN example]", "[JP example]" |

---

## [Main Section]

### Step 1: [Title]

\`\`\`bash
# Command example
\`\`\`

### Step 2: [Title]

[Instructions]

---

## Best Practices

\`\`\`
✅ GOOD: [Recommended pattern]
❌ BAD: [Anti-pattern]
\`\`\`

---

## Checklist

- [ ] [Requirement 1]
- [ ] [Requirement 2]
- [ ] [Requirement 3]
```

---

## Validation

```bash
# Check frontmatter
head -n 5 .claude/skills/[name]/SKILL.md

# Test trigger
# Ask Claude: "[trigger phrase]"
# Expected: Skill activates
```

---

## Checklist

- [ ] Directory created at `.claude/skills/[name]/`
- [ ] SKILL.md has valid frontmatter
- [ ] Name is kebab-case, no forbidden words
- [ ] Description has action + trigger
- [ ] Triggers cover EN and JP
- [ ] Steps are clear and actionable

Overview

This skill creates well-structured Claude Skills for the Miyabi project, producing a ready-to-use skill folder and manifest. It guides naming, triggers, inputs/outputs, and packaging so new skills follow project conventions and are automatable. Use it when you need a repeatable process to add or package capabilities for Claude agents.

How this skill works

The skill gathers requirements (purpose, triggers, input, output, tools), enforces naming rules, and generates a skill directory with a manifest, resources, and helper scripts. It validates frontmatter, trigger coverage (English + Japanese examples), and a clear step-by-step workflow so the skill activates reliably. It also provides a checklist for packaging and basic validation commands.

When to use it

Creating a new custom Claude Skill from a feature idea
Packaging an existing workflow into a reusable skill
Automating repetitive agent tasks or developer workflows
Onboarding contributors to ensure consistent skill structure
Preparing skills for parallel execution or CI integration

Best practices

Choose a kebab-case, descriptive name and avoid forbidden trademarks or protocol names
Write a one-line description with an action verb plus trigger conditions
Include clear triggers with both English and Japanese example phrases
Keep workflow steps short, actionable, and tool-specific (Bash, Read, Write, etc.)
Provide minimal example commands and a short checklist for validation

Example use cases

Generate a code-review skill that runs linters and summarizes issues when someone says "create skill" or "code review"
Package a deployment workflow into a skill that triggers on "package this workflow" and runs CI scripts
Build a test-generator skill that accepts input spec and outputs test cases and commands
Automate repo housekeeping tasks (branch cleanup, label management) as a reusable skill
Create a documentation extractor skill that collects reference docs into resources/ for the agent

FAQ

What naming rules must I follow?

Use kebab-case, lowercase, and descriptive names; avoid forbidden terms like company or protocol names.

What should the skill manifest include?

Include frontmatter with name, concise action-oriented description, allowed tools, and version plus a short purpose and triggers with examples.

How do I validate the skill?

Check the manifest frontmatter, confirm triggers cover English and Japanese examples, and run simple activation tests with a trigger phrase to ensure the skill activates.