creating-skills

name	creating-skills
description	Creates new Claude Code skills following best practices. Guides through skill structure, naming, descriptions, and progressive disclosure. Use when user wants to create a skill, build a skill, make a new capability, or asks about skill best practices.

Creating Skills

Guides creation of Claude Code skills using documented best practices.

Quick Start

For a new skill:

1. Ask user for skill purpose and target users
2. Generate using appropriate template
3. Validate against checklist

For reviewing existing skill:

1. Read SKILL.md and supporting files
2. Check against anti-patterns in reference.md
3. Report issues with fixes

Workflow: Create New Skill

Progress:
- [ ] Gather requirements (purpose, triggers, complexity)
- [ ] Choose template (basic or advanced)
- [ ] Generate skill structure
- [ ] Customize content
- [ ] Validate against checklist

Step 1: Gather Requirements

Ask user with AskUserQuestion:

• What should this skill do? (purpose)
• When should it activate? (trigger words)
• Simple or complex? (affects structure)

Step 2: Choose Structure

Complexity	Structure	When to Use
Basic	Single SKILL.md	Simple workflows, < 200 lines
Advanced	SKILL.md + references	Complex domains, multiple workflows

Step 3: Generate Skill

Use templates from templates/ directory:

• templates/basic.md - Single file skill
• templates/advanced.md - Multi-file with references

Create in appropriate location:

• ~/.claude/skills/ - Personal skills
• .claude/skills/ - Project skills (git-tracked)

Step 4: Validate

Run through checklist before finishing:

Validation Checklist:
- [ ] Name: gerund format (verb-ing), lowercase, hyphens only
- [ ] Name: max 64 characters, no reserved words
- [ ] Description: explains WHAT and WHEN to use
- [ ] Description: third person, max 1024 characters
- [ ] SKILL.md: under 500 lines (ideally under 300)
- [ ] References: one level deep from SKILL.md
- [ ] No time-sensitive info (dates, versions)
- [ ] Consistent terminology throughout
- [ ] Examples: concrete input/output pairs
- [ ] Paths: Unix-style only (forward slashes)

Naming Rules

Format: verb-ing-noun (gerund form)

• processing-pdfs
• analyzing-spreadsheets
• managing-databases

Constraints:

• Max 64 characters
• Lowercase letters, numbers, hyphens only
• No: anthropic, claude, XML tags

Description Best Practices

The description determines when Claude activates the skill.

Include:

• What the skill does
• When to use it
• Trigger words users might say

Format: Third person, active voice

# Good
description: |
  Extracts text and tables from PDF files. Use when working with
  PDFs, forms, or document extraction.

# Bad - first person
description: I help you process PDF files

Progressive Disclosure

Keep SKILL.md lean. Move details to reference files:

skill-name/
├── SKILL.md          # Overview, workflow (always loaded)
├── reference.md      # Detailed guidelines (loaded when needed)
├── examples.md       # Extended examples
└── templates/        # Reusable templates

Key rule: References should be one level deep. All reference files link directly from SKILL.md.

See reference.md for detailed best practices and anti-patterns.

Frontmatter Schema

---
name: skill-name-here
description: |
  What it does and when to use it. Include trigger words.
allowed-tools: Read, Grep  # Optional: restrict available tools
---

Anti-Patterns (Quick Reference)

Avoid	Do Instead
Windows paths scripts\file.py	Unix paths scripts/file.py
Too many options	Provide sensible default
Assuming tools installed	List required packages
Vague descriptions	Specific with trigger words
Deeply nested references	One level deep
Time-sensitive info	Avoid dates/versions

Full anti-patterns guide: reference.md

Example: Basic Skill

---
name: formatting-markdown
description: |
  Formats and lints Markdown files. Use when user mentions
  markdown formatting, MD files, or document styling.
---

# Formatting Markdown

## Workflow

1. Read target file
2. Apply formatting rules
3. Report changes

## Rules

- Headers: ATX style (#)
- Lists: consistent markers
- Code blocks: fenced with language