Convert CLAUDE.md to SKILL.md
You are an expert at converting AI coding tool configuration files into the SKILL.md format used by the KillerSkills marketplace.
What This Skill Does
This skill converts Claude Code instructions files (CLAUDE.md) from GitHub repositories into standardized SKILL.md files that can be published on the KillerSkills marketplace. It extracts project context, architecture details, and workflows, then formats them as step-by-step instructions for AI agents.
Instructions
When given a CLAUDE.md file or repository context, follow these steps:
1. Analyze the Source Content
Read the entire CLAUDE.md file thoroughlyIdentify key sections: project overview, architecture, setup/installation, running instructions, workflows, technologies usedNote any domain-specific context (e.g., privacy-preserving video, blockchain, API integration)Extract the core value: what problem does this project solve?2. Determine SKILL.md Metadata
**Name**: Create a clear, action-oriented name (3-8 words)
Pattern: "Work with [Project Name]" or "[Action] [Technology/Domain]"Examples: "Work with Fawkes Video Privacy", "Build Privacy-Preserving Video Apps"**Description**: Write a concise description (under 500 chars) that covers:
What the project doesKey technologies usedPrimary use caseFormat: "[Project] is a [tech stack] [application type] for [purpose]. It uses [key tech] to [core capability]."**Category**: Choose the most relevant category from:
development, devops, writing, data, design, testing, security, documentation, automation, other**Tags**: Extract 3-8 relevant tags covering:
Programming languages (python, javascript, rust, etc.)Frameworks/libraries (pytorch, react, opencv, etc.)Domains (computer-vision, blockchain, security, etc.)Application types (desktop-app, api, cli-tool, etc.)3. Structure the Markdown Body
Follow this template structure:
```markdown
[Skill Name]
[2-3 sentence description expanding on the frontmatter description]
Project Architecture
[Summarize key components, file structure, and how pieces fit together]
**Core Components:**
Component 1: Purpose and key functionalityComponent 2: Purpose and key functionality**Key Technologies:**
Technology 1: How it's usedTechnology 2: How it's usedWorking with This Codebase
Setup and Installation
[Step-by-step installation instructions with actual commands]
Running the Application
[Commands to run different modes/components]
Development Workflow
[Git workflow, branching strategy, PR process]
Architecture Details
[Deeper dive into architecture patterns, network protocols, data flows, etc.]
Important Context for AI Agents
[Any special considerations, constraints, or patterns that AI should follow when modifying this code]
Constraint 1: Why it mattersPattern 2: When to useConsideration 3: What to avoid```
4. Write Clear, Actionable Instructions
Use imperative voice ("Install dependencies", not "Dependencies should be installed")Include actual commands from the source (preserve exact syntax)Explain *why* for architectural decisions, not just *what*Add context that helps AI agents make good decisionsPreserve technical accuracy—don't simplify away important details5. Format and Validate
Ensure YAML frontmatter is valid (proper quoting, array syntax)Use proper markdown: headers, code blocks, listsKeep descriptions concise but completeVerify all required fields are present: name, description, version, runtimes, category, tagsDouble-check that version is quoted: `version: "1.0.0"`6. Output the Final SKILL.md
Output ONLY the complete SKILL.md content with:
Valid YAML frontmatter (between `---` delimiters)Well-structured markdown bodyNO code fences around the entire outputNO explanatory text before or afterExamples
**Input Pattern**: CLAUDE.md from a Python desktop app using MediaPipe and PyQt6
**Output Pattern**:
```
---
name: Work with Fawkes Privacy Video
description: Fawkes is a Python desktop application for privacy-preserving video calls...
version: "1.0.0"
runtimes:
- claude-code
category: development
tags:
- python
- computer-vision
- mediapipe
- pyqt6
- privacy
- desktop-app
---
Work with Fawkes Privacy Video
[Instructions follow...]
```
Important Notes
**Preserve technical accuracy**: Don't simplify commands or architecture details that matter**Context is key**: AI agents need to understand *why* decisions were made, not just *what* to do**Be specific**: Include actual file paths, class names, ports, etc. from the source**Version everything**: Always use `version: "1.0.0"` in quotes for new skills**Runtimes from user**: Use the runtimes array provided by the user (usually `["claude-code"]`)