Create Skill

Create Skill Command

Purpose

Interactive wizard for creating skills - specialized, reusable workflows invoked by agents. Ensures skills follow standards and integrate properly.

⚙️ Configuration

Usage

/create-skill [options]

Options

• --name <name>: Skill name (kebab-case)
• --category <category>: Skill category
• --interactive: Full interactive mode (default)
• --template <type>: Use template

When to Use

• Reusable workflows needed
• Domain expertise codification
• Complex procedures documentation
• Best practice standardization

Skill Creation Process

Step 1: Discovery & Planning

Wizard Questions:

1. Skill name (kebab-case)
2. Skill category
3. One-line description
4. Detailed purpose
5. Primary users (agents)
6. Skill type

Step 2: Workflow Definition

Define:

1. Input requirements
2. Process steps
3. Output/deliverables
4. Success criteria

Step 3: File Generation

File Created: {base_path}/{category}/{skill-name}.md

YAML Frontmatter:

---
name: {skill-name}
category: {category}
description: {description}
usage: {usage_context}
input: {input_requirements}
output: {output_produced}
---

Structure:

# {Skill Name}

## Overview
{purpose_and_users}

## When to Use
{scenarios}

## Prerequisites
{requirements}

## Workflow
### Step 1: {name}
{actions_and_validation}

## Output
{deliverables_and_criteria}

## Examples
{scenarios}

## Error Handling
{errors_and_solutions}

Step 4: Integration & Documentation

Updates:

• Skills README
• Agent documentation (if agent-specific)
• Quick start (if widely used)

Step 5: Validation

Checks:

• Name follows conventions
• YAML valid
• All sections complete
• Workflow clear and actionable
• Prerequisites defined
• Success criteria measurable
• File in correct directory

Test: Invoke from agent context

Step 6: Commit

Format:

feat(skills): add {skill-name} skill

- {description}
- Primary users: {agents}

Workflow: {key_steps}
Output: {deliverables}

Updates:
- {skill_file} (new)
- README.md (updated)

Interactive Wizard Flow

🎯 Create New Skill Wizard

📝 Step 1: Skill Identity
Skill Name: {input}
Category: {selection}
Description: {input}
Type: {selection}

📋 Step 2: Purpose & Users
Purpose: {input}
Primary users: {input}

🔧 Step 3: Workflow Definition
Steps: {input}
Input: {input}
Output: {input}
Success criteria: {input}

📝 Step 4: Review & Confirm
{summary}
Proceed? (y/n)

✨ Creating Skill
✓ Generated file
✓ Updated documentation
✓ Validation passed

Skill Templates

Workflow Skill

---
name: {name}
category: {category}
description: {desc}
usage: Invoke when {context}
input: {requirements}
output: {deliverables}
---

# {Name}

## Overview
**Purpose**: {purpose}
**Primary Users**: {agents}

## Workflow
### Step 1: {name}
**Objective**: {goal}
**Actions**: {steps}
**Validation**: {checks}
**Output**: {result}

## Output
**Produces**: {outputs}
**Success Criteria**: {criteria}

Validation Skill

---
name: {name}
category: validation
description: {desc}
usage: Use to validate {what}
input: {what_to_validate}
output: Validation report
---

# {Name}

## Validation Checklist
### {Category}
- [ ] {check}

## Validation Process
{steps}

## Output Format
{report_format}

Validation Rules

Skill Name

• Format: kebab-case
• Length: 3-40 characters
• Pattern: Configured regex
• Must be unique

YAML Frontmatter

Required: name, category, description, usage, input, output

Valid Categories: Configured categories

Valid Types: Configured types

Directory Structure

{base_path}/
├── testing/
├── development/
├── validation/
├── documentation/
├── automation/
└── analysis/

Best Practices

Design

• Generic and reusable
• Parameterized for context
• Modular steps
• Clear objectives
• Actionable steps
• Validated at each step

Documentation

• Realistic examples
• Error scenarios documented
• Clear prerequisites
• Agent-aware design

Common Patterns

Testing Workflow

1. Prerequisites check
2. Test planning
3. Test implementation
4. Validation
5. Documentation

Validation Workflow

1. Input validation
2. Criteria checking
3. Result compilation
4. Report generation

Related Commands

• /create-agent - Create new agent
• /create-command - Create new command
• /help - System help

Notes

• Skills invoked by agents, not users
• Keep focused on single responsibility
• Comprehensive docs critical
• Test from actual agent context
• Skills evolve with patterns
• Use action-oriented names