Agent Skill Generator

Bash
# Create skill structure
mkdir my-skill-name
cd my-skill-name
touch SKILL.md README.md
mkdir examples tests

Progress:

• Parse skill requirements
• Generate kebab-case skill name
• Create folder structure
• Write SKILL.md with proper format
• Add examples and tests
• Provide improvement suggestions

Step-by-step Process:

1. Extract Requirements

   • Identify the core capability
   • Determine trigger scenarios
   • Note any specific tools/technologies

2. Generate Structure

   skill-name/
   ├── SKILL.md          # Main skill file
   ├── README.md         # Installation/usage
   ├── examples/         # Sample inputs/outputs
   ├── tests/           # Test cases
   └── assets/          # Supporting files
   

3. Create SKILL.md

   • YAML frontmatter (name, description)
   • Quick Start section
   • Workflow with checklist
   • Examples with input/output
   • Best Practices
   • Common Pitfalls

4. Validate Format

   • Name: gerund form, kebab-case, <64 chars
   • Description: third person, includes triggers, <1024 chars
   • Content: ~500 lines max, concrete examples

Example 1: Input: "I want a skill for analyzing CSV data and generating reports" Output:

analyzing-csv-data/
├── SKILL.md (with pandas workflow, visualization examples)
├── examples/sample.csv
├── examples/output_report.html
└── README.md

Example 2: Input: "Create a skill for writing technical documentation" Output:

writing-technical-docs/
├── SKILL.md (with doc structure, templates)
├── examples/api_doc_template.md
├── examples/user_guide_sample.md
└── tests/doc_quality_checklist.md

• Use descriptive gerund names: processing-images not image-tool
• Include concrete examples: Real code snippets, not pseudocode
• Add progress checklists: For multi-step workflows
• Provide default configurations: Don't overwhelm with options
• Test the workflow: Ensure examples actually work

• Vague descriptions: "Helps with data" → "Analyzes CSV files and generates statistical reports"
• Missing trigger phrases: Always include "Use when..." scenarios
• Too many options: Give one recommended approach, not five alternatives
• Abstract examples: Use real data/code, not placeholder text
• Overly complex: If >500 lines, suggest splitting into multiple skills

After generating a skill, provide:

1. Enhancement ideas: Additional features to consider
2. Integration opportunities: How it connects with other skills
3. Testing recommendations: Key scenarios to validate
4. Documentation gaps: Areas needing more detail