Contributing to Claude Skills Marketplace

Contributing to Claude Skills Marketplace

Thank you for your interest in contributing! This guide will help you create high-quality Skills that work reliably across different Claude models.

Skill Structure

A basic Skill consists of:

skill-name/
├── SKILL.md          # Required: Main skill file
├── reference.md      # Optional: Detailed reference material
├── examples/         # Optional: Example files or templates
└── scripts/          # Optional: Executable scripts

SKILL.md Format

---
name: skill-name
description: What it does and when to use it. Include activation triggers.
model: sonnet              # Optional: Preferred model (sonnet, opus, haiku)
---

# Skill Title

Brief overview (1-2 sentences).

## When to Use

Specific activation scenarios:
- User says "X"
- Context includes Y
- Task involves Z

## Instructions

Step-by-step instructions for Claude...

Best Practices

1. Write Clear Descriptions

**Good:**

description: Analyze vendorlist extraction performance and F1-scores. Use when user mentions vendorlist evaluation, low F1-scores, extraction accuracy issues, or asks to analyze vendorlist performance.

**Bad:**

description: Helps with vendorlists

The description should answer:

• What does this skill do?
• When should it activate?
• Why would someone use it?

2. Use Gerund Names

**Good:** `git-pushing`, `test-fixing`, `review-implementing`

**Bad:** `git-push`, `fix-test`, `implement-review`

Gerund forms (-ing) indicate ongoing processes and match natural language patterns.

3. Keep Skills Focused

Each Skill should do ONE thing well. If you find yourself writing "and also" in the description, consider splitting into separate Skills.

4. Structure for Progressive Disclosure

**Keep SKILL.md under 500 lines:**

• Core instructions in SKILL.md
• Detailed reference material in reference.md
• Examples in separate files

This allows Claude to load only what's needed.

5. Assume Claude's Knowledge

Don't explain basic concepts. Claude already knows Python, git, testing frameworks, etc.

**Good:**

Run `pytest -k "pattern"` to test specific cases.

**Bad:**

Pytest is a testing framework for Python. The -k flag filters tests by pattern. You should use this command: `pytest -k "pattern"` to run tests that match the pattern.

6. Match Specificity to Risk

**High-risk operations** (database migrations, deployments):

• Provide step-by-step instructions
• Include validation checks
• Require confirmation

**Low-risk operations** (code formatting, documentation):

• Give general guidance
• Allow flexibility

7. Use Platform-Agnostic Paths

**Good:** `backend/src/utils/helper.py`

**Bad:** `backend\src\utils\helper.py`

Always use forward slashes for file paths.

8. Test Across Models

Test your Skill with:

• Claude Sonnet (balanced)
• Claude Opus (most capable)
• Claude Haiku (fastest)

Some models may need more det