Skill Development

Skill Development

Guide for authoring high-quality, cross-platform agent skills.

SKILL.md Format

Every skill is a markdown file with YAML frontmatter:

---
name: my-skill-name
description: Use when the user asks to "do X", "perform Y", or needs help with Z. Provides step-by-step procedures for accomplishing the task.
---

# My Skill Name

## Purpose
What this skill enables the agent to do.

## Procedures
Step-by-step instructions the agent follows.

## When to Load
- Load when: [trigger conditions]
- Not needed when: [exclusion conditions]

Frontmatter Spec (agentskills.io)

Required Fields

Optional Fields

Name Rules

• Lowercase letters, digits, hyphens only
• Cannot start or end with a hyphen
• Max 64 characters
• Must match the parent directory name exactly

Directory Structure

skill-name/
├── SKILL.md              # Required: main skill file
├── references/           # Optional: supporting documents
│   ├── api-patterns.md
│   └── common-errors.md
└── examples/             # Optional: usage examples
    └── basic-setup.md

Reference files enable progressive disclosure — the main SKILL.md stays concise while references/ holds deep-dive content the agent loads on demand.

See handbook/agentskills-io-spec.md for the full specification reference.

Description Best Practices

The description field is the primary trigger mechanism. Agents load skills based on keyword matching against descriptions.

DO

• Start with "Use when..." or "This skill should be used when..."
• Include quoted trigger phrases: "deploy to production", "fix auth bug"
• List concrete user prompts that should activate it
• Include negative triggers: "Not needed when..."

DON'T

• Write vague descriptions: "Helps with development"
• Omit trigger phrases — the skill won't be discoverable
• Make descriptions longer than 1024 characters
• Duplicate another skill's trigger phrases

Example (good)

description: Use when the user asks to "set up authentication", "add login", "implement JWT", "secure API endpoints", or needs guidance on session management, OAuth flows, or password hashing. Not needed for basic CRUD operations or frontend styling.

Example (bad)

description: Authentication skill for web applications.

Testing a Skill Locally

Before publishing, verify your skill works:

1. Place it: Copy skill-name/SKILL.md to .claude/skills/skill-name/SKILL.md
2. Validate frontmatter:

   python3 -c "
   import yaml
   content = open('.claude/skills/skill-name/SKILL.md').read()
   fm = yaml.safe_load(content.split('---', 2)[1])
   assert fm.get('name') == 'skill-name', 'Name mismatch'
   assert fm.get('description'), 'Missing description'
   print('PASS')
   "
   

3. Test trigger: Start a new session and type a prompt that should activate the skill
4. Verify loading: The agent should load and follow the skill's procedures

Publishing

See handbook/publishing-skills.md for the full publishing guide. Quick version:

1. Create a public GitHub repo
2. Place skills in skills/<name>/SKILL.md
3. Push — skills.sh auto-indexes it
4. Verify: npx skills find <your-skill-name>

When to Load This Skill

Load skill-development when:

• Creating a new skill from scratch
• Improving a skill's description or trigger phrases
• Structuring a skill directory with references and examples
• Validating frontmatter format
• Preparing a skill for publishing

Not needed when:

• Installing existing skills (use skill-management)
• Searching for skills (use /skills-search)
• Removing skills (use skill-management)