Skill Creator

Create production-ready Claude Code skills with proper structure, frontmatter, and documentation.

Prerequisites

• Understanding of the target skill's purpose and scope
• Knowledge of relevant tools and APIs the skill will use
• Clear idea of trigger phrases users will say

Workflow

Step 1: Gather Requirements

Ask the user for the following information:

1. Skill name - concise, descriptive (e.g., "Code Reviewer", "Git Workflow")
2. Purpose - what problem does this skill solve?
3. Category - devtools, productivity, research, creative, etc.
4. Trigger phrases - 3-7 natural language phrases users would say
5. Tools needed - which Claude Code tools will it use (Bash, Read, Write, Edit, etc.)
6. Inputs - what information does the skill need from the user
7. Outputs - what does the skill produce

Step 2: Design the Workflow

Break the skill into sequential steps:

1. Map out the logical flow from input to output
2. Identify decision points and branching logic
3. Determine which tools are called at each step
4. Define error handling for each step
5. Plan validation checkpoints

Step 3: Write the Frontmatter

Generate valid YAML frontmatter with these required fields:

---
name: <Skill Name>
description: |
  <2-3 sentence description of what the skill does.>
  Trigger phrases: "<phrase1>", "<phrase2>", "<phrase3>"
version: 1.0.0
author: <author>
tags: [<tag1>, <tag2>, <tag3>]
---

Step 4: Write the Skill Body

Structure the SKILL.md with these sections:

1. Title - H1 heading matching the skill name
2. Overview - 1-2 paragraphs explaining the skill
3. Prerequisites - what must be in place before running
4. Workflow - numbered steps with detailed instructions
5. Output Format - exact format of what gets produced
6. Examples - 2-3 concrete usage examples
7. Best Practices - tips for effective use
8. Error Handling - common issues and recovery steps

Step 5: Validate the Skill

Check the generated SKILL.md against these criteria:

• Frontmatter parses as valid YAML
• Description includes trigger phrases
• All workflow steps are actionable
• Output format is clearly defined
• Examples cover common use cases
• Error handling addresses likely failures
• File is between 80-250 lines
• No ambiguous instructions

Step 6: Write the File

Use the Write tool to create the SKILL.md file at the correct path:

skills/<category>/<skill-name>/SKILL.md

Output Format

The skill produces a complete SKILL.md file with:

• Valid YAML frontmatter block
• Markdown body with all required sections
• Consistent heading hierarchy (H1 > H2 > H3)
• Code blocks for any templates or examples

Examples

Example 1: Creating a Code Review Skill

User: "Create a skill that reviews pull requests"

Action: Generate a SKILL.md that walks through reading diff output, analyzing code quality, checking for common issues, and producing a structured review comment.

Example 2: Creating a Documentation Skill

User: "Build a skill for generating API docs"

Action: Generate a SKILL.md that scans source files, extracts function signatures, reads docstrings, and produces formatted API documentation.

Example 3: Creating a Deployment Skill

User: "Make a skill for deploying to production"

Action: Generate a SKILL.md that runs pre-deploy checks, builds artifacts, executes deployment commands, and verifies the deployment succeeded.

Best Practices

1. Keep trigger phrases natural - use phrases people actually say, not technical jargon
2. Be specific in workflow steps - each step should have a clear, verifiable outcome
3. Include error recovery - every tool call can fail; plan for it
4. Test with edge cases - consider empty inputs, large files, missing dependencies
5. Version your skills - use semantic versioning in frontmatter
6. One responsibility per skill - avoid skills that try to do everything
7. Document assumptions - state what the skill expects about the environment
8. Use consistent formatting - follow the same Markdown conventions throughout

Error Handling