Skill Creator

This skill creates other skills (meta!).

About

The skill-creator skill guides you through the complete process of creating a new Timus skill, from understanding requirements to packaging for distribution.

When to Use This Skill

Use this skill when:

• User wants to create a new skill
• User needs help structuring a skill
• User wants to add scripts/references to a skill
• User needs to package a skill for distribution
• User asks about skill best practices

Skill Creation Process

Step 1: Understand with Concrete Examples

Goal: Clearly understand what the skill should do.

Ask the user:

1. "What should this skill do?"
2. "Can you give me 3 concrete examples of how it would be used?"
3. "What would a user say that should trigger this skill?"

Don't overwhelm: Ask 1-2 questions at a time, follow up as needed.

Example:

User: "Create a skill for PDF processing"
→ "What should it do? Extract text, merge PDFs, rotate pages?"
→ "Example 1: 'Extract text from this PDF'"
→ "Example 2: 'Merge these PDFs into one'"

Step 2: Plan Resources

Analyze the concrete examples and plan resources:

Scripts (Python/Bash):

• When: Code is rewritten repeatedly OR deterministic reliability needed
• Example: scripts/rotate_pdf.py for PDF rotation
• Include: If same code pattern appears multiple times

References (Documentation):

• When: Schemas, API docs, policies needed
• Example: references/schema.md for database structure
• Include: Detailed docs, schemas, workflows, examples

Assets (Templates, Images):

• When: Boilerplate or resources needed for output
• Example: assets/template.html for web projects
• Include: Templates, fonts, images, sample files

Step 3: Initialize Skill

Tool: init_skill_tool

Usage:

{
  "method": "init_skill_tool",
  "params": {
    "name": "pdf-processor",
    "description": "Process PDF files for text extraction, rotation, and merging",
    "resources": ["scripts", "references"],
    "examples": true
  }
}

This creates:

skills/pdf-processor/
├── SKILL.md (with template)
├── scripts/ (with example.py if examples=true)
└── references/ (with REFERENCE.md if examples=true)

Step 4: Edit SKILL.md

Frontmatter (YAML):

---
name: pdf-processor
description: Process PDF files for: (1) text extraction, (2) rotation, 
             (3) merging. Use when working with PDF files.
---

Key Rule: Description goes in frontmatter, NOT in body!

• Frontmatter: Always loaded → Perfect for trigger keywords
• Body: Only on trigger → Don't put "When to Use" here

Body (Markdown):

# PDF Processor

## Quick Start
Extract text with pdfplumber:
```python
import pdfplumber
with pdfplumber.open("doc.pdf") as pdf:
    text = pdf.pages[0].extract_text()

Advanced Features

• Form filling: See FORMS.md
• API Reference: See API.md

**Writing Guidelines:**
- Always use imperative/infinitive form
- ✅ "Extract the text"
- ❌ "You should extract..."
- Keep under 500 lines
- Be concise - context window is limited!

### Step 5: Add Scripts & References

**Create scripts:**
```python
# skills/pdf-processor/scripts/rotate_pdf.py
import PyPDF2

def rotate_pdf(input_path, output_path, rotation):
    with open(input_path, 'rb') as f:
        reader = PyPDF2.PdfReader(f)
        writer = PyPDF2.PdfWriter()
        
        for page in reader.pages:
            page.rotate(rotation)
            writer.add_page(page)
        
        with open(output_path, 'wb') as out:
            writer.write(out)

if __name__ == "__main__":
    import sys
    rotate_pdf(sys.argv[1], sys.argv[2], int(sys.argv[3]))

Test scripts:

• Actually run them before bundling
• Ensure no bugs
• Output matches expectations

Create references:

# skills/pdf-processor/references/PDFLIB.md

## PyPDF2 API

### PdfReader
- `pages` - List of pages
- `getPage(n)` - Get specific page

### Page Methods
- `rotateClockwise(degrees)` - Rotate page
- `extract_text()` - Extract text content

Step 6: Package Skill

Tool: package_skill_tool (coming soon)

For now:

• Validate: Check SKILL.md structure
• Test: Run on real examples
• Distribute: Copy skill folder

Validation Checklist

Before completing, verify:

• YAML frontmatter valid?
• Name between 2-64 chars?
• Description > 10 chars, < 500 chars?
• Body > 50 chars?
• Scripts tested?
• References linked from SKILL.md?
• No extraneous files (README, etc.)?

Progressive Disclosure

This skill uses the three-level system:

1. Metadata (this frontmatter) - Always loaded (~100 words)
2. SKILL.md Body - Loaded on trigger (this document, <5k words)
3. References - Loaded on demand:
   • best-practices.md - OpenClaw best practices
   • workflow-patterns.md - Common workflow patterns

References

• Best Practices: See best-practices.md
• Workflow Patterns: See workflow-patterns.md