Claude Skills Authoring Guide

Overview

This skill provides best practices for creating effective Claude Agent Skills that are discoverable, concise, and well-tested.

When to Use

• Creating new skills: Before writing a new SKILL.md
• Improving existing skills: When refactoring or enhancing skills
• Reviewing skill quality: For code review of skill files
• Learning skill authoring: Understanding the skill architecture

Core Principles

1. Concise is Key

The context window is a shared resource. Challenge each piece of information:

• "Does Claude really need this explanation?"
• "Can I assume Claude knows this?"
• "Does this paragraph justify its token cost?"

Default assumption: Claude is already very smart. Only add context Claude doesn't already have.

2. Progressive Disclosure

SKILL.md serves as an overview that points to detailed materials as needed:

• Keep SKILL.md body under 500 lines
• Split content into separate files when approaching this limit
• Claude loads additional files only when needed

3. Appropriate Degrees of Freedom

Match specificity to task fragility:

Quick Reference

YAML Frontmatter Requirements

---
name: skill-name        # lowercase, hyphens, max 64 chars
description: Describes what it does and when to use it.  # max 1024 chars
---

Naming convention: Use gerund form (verb + -ing)

• Good: processing-pdfs, analyzing-data, testing-code
• Avoid: helper, utils, tools

Description rules:

• Always write in third person
• Include what the skill does AND when to use it
• Be specific and include key terms for discovery

See NAMING.md for detailed naming guidelines.

File Structure

my-skill/
├── SKILL.md              # Main instructions (loaded when triggered)
├── REFERENCE.md          # API reference (loaded as needed)
├── EXAMPLES.md           # Usage examples (loaded as needed)
└── scripts/
    └── utility.py        # Executed, not loaded into context

See STRUCTURE.md for progressive disclosure patterns.

Skill Creation Workflow

Step 1: Identify the Gap

Before writing documentation, identify what Claude struggles with:

1. Run Claude on representative tasks without a skill
2. Document specific failures or missing context
3. Create 3+ evaluation scenarios

Step 2: Write Minimal Instructions

Create just enough content to address the gaps:

---
name: my-skill
description: [What it does]. Use when [trigger conditions].
---

# [Skill Title]

## Quick Start
[Minimal working example]

## Advanced Features
See [REFERENCE.md](REFERENCE.md) for details.

Step 3: Test and Iterate

1. Test with real usage scenarios
2. Observe Claude's navigation patterns
3. Refine based on failures

See WORKFLOWS.md for detailed development workflow.

Common Patterns

Template Pattern

Provide output format templates:

## Report Structure

Use this template:

```markdown
# [Title]

## Summary
[One-paragraph overview]

## Key Findings
- Finding 1
- Finding 2
```

Examples Pattern

Show input/output pairs:

## Commit Message Format

**Input**: Added user authentication
**Output**:
```
feat(auth): implement authentication

Add login endpoint and token validation
```

Conditional Workflow Pattern

Guide through decision points:

## Workflow

1. Determine task type:
   - **Creating new?** → See "Creation workflow"
   - **Editing existing?** → See "Editing workflow"

Anti-Patterns to Avoid

Detailed Documentation

• NAMING.md: Naming conventions and description guidelines
• STRUCTURE.md: File organization and progressive disclosure
• WORKFLOWS.md: Development workflow and iteration
• CHECKLIST.md: Quality checklist before sharing

Related Skills

• writing-technical-docs: General documentation principles
• applying-solid-principles: Code quality for utility scripts