code

What are Agent Skills?

Official Claude Code feature by Anthropic

Agent Skills are an official feature of Claude Code (Anthropic's AI coding assistant) that extend Claude's capabilities with specialized knowledge, workflows, and instructions. Think of them as "expert modules" that give Claude domain-specific expertise on demand.

psychologyHow Agent Skills Work

Skills are markdown files (typically named skill.md) that contain:

  • check_circleYAML frontmatter with metadata (name, description, allowed tools)
  • check_circleInstructions that guide Claude's behavior
  • check_circleExamples of how to apply the skill
  • check_circleCommands & workflows for common tasks

When you invoke a skill, Claude Code loads the skill's instructions into context, giving Claude specialized knowledge for that domain. Skills can be invoked automatically based on task context or manually via slash commands like /skill.

devicesCustom Instructions Across AI Tools

The concept of giving an AI coding assistant specialized instructions exists across all major tools. Here's how each one handles it:

Claude Code uses Skills — modular markdown files with YAML frontmatter. Place them in .claude/skills/ (project) or ~/.claude/skills/ (global). Invoke with the /skill slash command.

description.claude/skills/tdd.md
---
name: test-driven-development
description: Enforce strict TDD workflow
allowed-tools: Read, Write, Edit, Bash
---

# TDD Workflow

## Red Phase
1. Write ONE failing test
2. Run it - confirm it fails

## Green Phase
1. Write MINIMAL code to pass
2. Run tests - confirm passing

## Refactor Phase
1. Improve code structure
2. Keep tests passing

descriptionExample skill.md File

descriptionskill.md
---
name: test-driven-development
description: Enforce strict TDD workflow with red-green-refactor cycle.
allowed-tools: Read, Write, Edit, Bash
---

# Test-Driven Development

Follow the TDD cycle strictly for all implementations.

## The TDD Cycle

### 🔴 RED Phase
1. Write ONE failing test
2. Run it - confirm it fails
3. Verify the failure message is meaningful

### 🟢 GREEN Phase
1. Write MINIMAL code to pass
2. No more than necessary
3. Run tests - confirm passing

### 🔵 REFACTOR Phase
1. Improve code structure
2. Remove duplication
3. Keep tests passing
4. Commit changes

## Rules
- Never write production code without a failing test
- Never write more test than needed to fail
- Never write more code than needed to pass

build_circleHow to Create a Agent Skill

1

Create a skill.md file

Create a markdown file with YAML frontmatter containing the skill name, description, and optionally which tools the skill can use (Read, Write, Edit, Bash, etc.).

2

Write clear instructions

Document the workflow, rules, and examples. Be specific about what Claude should and shouldn't do. Include code examples and command-line instructions where relevant.

3

Install and test locally

Place the skill in your project's .claude/skills/ directory for project-specific use, or ~/.claude/skills/ for global availability. Use the /skill slash command to load and test it.

4

Share on Mirdan.ai

Once your skill is working well, share it with the community! Submit it to Mirdan.ai so others can benefit from your work.

add_circleSubmit Your Skill

menu_bookOfficial Anthropic Documentation

Skills are an official feature of Claude Code by Anthropic. For the most up-to-date information on skill syntax, capabilities, and best practices, refer to the official documentation:

open_in_newClaude Code Documentationopen_in_newSkills Reference Guide

Ready to explore Agent Skills?

Browse our curated collection of community and official skills for Claude Code.

exploreBrowse Skills