1.9 KiB
1.9 KiB
Skills Reference
Quick reference for creating agent skills.
Structure
my-skill/
├── SKILL.md (required)
├── reference.md (optional)
├── examples.md (optional)
├── scripts/
│ └── helper.py
└── templates/
└── template.txt
Location
- Project:
.claude/skills/skill-name/ - Personal:
~/.claude/skills/skill-name/
SKILL.md Format
---
name: my-skill-name
description: What it does and when to use it. Include trigger keywords.
allowed-tools: Read, Write, Bash
---
# My Skill
## Instructions
1. Clear step-by-step guidance
2. What to do and how
## Examples
Simple, minimal examples
Description Tips
Include WHEN to use:
- ✅ "Extract PDF text. Use when working with PDF files."
- ❌ "Extract PDF text."
Tool Access
- Omit
allowed-tools→ inherit all tools - Specify tools → restrict access (e.g., read-only:
Read, Grep, Glob)
Progressive Disclosure
Reference other files:
For advanced usage, see [reference.md](reference.md)
Run helper: `python scripts/helper.py`
Claude loads files only when needed.
Simple Examples
Code reviewer:
---
name: code-reviewer
description: Review code quality. Use after writing or modifying code.
allowed-tools: Read, Grep, Glob
---
# Code Reviewer
## Checklist
- Readable code
- Error handling
- No duplicated code
- Security issues
Review and provide specific feedback.
Commit helper:
---
name: commit-helper
description: Generate commit messages. Use when staging commits.
allowed-tools: Bash(git diff:*), Bash(git status:*)
---
# Commit Helper
## Instructions
1. Run `git diff --staged`
2. Generate clear commit message:
- Summary under 50 chars
- Present tense
- Explain what and why
Tips
- One skill = one focused capability
- Clear, specific descriptions
- Simple instructions
- Test with relevant requests