🌍 Language: English | Português | Español

From prompts to systems — learn how to build reusable AI skills.

Skills are one of the core building blocks of modern AI systems. They allow you to define reusable, task-specific capabilities that an AI system can automatically apply when relevant — without repeating instructions every time.

⭐ If this project helped you, please consider giving it a star — it helps others discover it and supports the work!

• Skills Quick Guide
• What are Skills
• How Skills Work
• Skill Structure
• Where Skills Live
• Skills vs Other Customizations
• When to Use Skills
• Example Use Cases
• Creating Your First Skill
• Best Practices
• Common Mistakes
• Advanced Concepts
• Why Skills Matter
• Resources
• Final Thought
• Contributing
• Connect with me

A quick visual reference to understand how skills work — from structure to activation.

💡 Use this as a quick reference while building your own skills.

A skill is a structured set of instructions that teaches an AI system how to perform a specific task.

In Claude Code, a skill is:

• A directory
• Containing a SKILL.md file
• With metadata and instructions

Once defined, the system can discover and use the skill automatically.

Instead of repeating prompts, you encode behavior once — and reuse it.

Skills follow a semantic matching workflow:

1. The system loads only skill names and descriptions
2. A user request is received
3. The system compares the request with available skill descriptions
4. Matching skills are selected
5. The full skill content is loaded and executed

This means:

• Skills are loaded on demand
• They do not clutter context unnecessarily
• They activate only when relevant

Each skill lives inside a directory and must contain a SKILL.md file.

---
name: pr-review
description: Reviews pull requests for code quality. Use when reviewing PRs or checking code changes.
---

Below the frontmatter, you define the instructions:

When reviewing a PR:

1. Analyze code changes
2. Check for best practices
3. Suggest improvements
4. Format output as structured feedback

Skills can be scoped at different levels:

• Location: ~/.claude/skills
• Available across all your projects
• Ideal for personal workflows

• Location: .claude/skills (inside a repository)
• Shared via version control
• Used for team standards

👉 Skills are automatic and contextual

Use a skill when:

• You repeat the same instructions frequently
• You need consistent output formats
• You want reusable workflows
• You want to share knowledge across a team

Rule of thumb:

If you explain the same thing more than once, it should be a skill.

• Code review standards
• Commit message formatting
• Documentation templates
• Debugging checklists
• Data transformation workflows
• Clinical text anonymization

👉 Explore ready-to-use skills: examples

Some examples are inspired by existing open-source skill repositories.

mkdir -p ~/.claude/skills/my-skill

Create a SKILL.md file inside your skill directory, or copy from the provided template.

---
name: my-skill
description: Describe what the skill does and when to use it.
---

Define:

• Steps
• Rules
• Output format

• Write clear and specific descriptions (this controls activation)
• Keep instructions structured and explicit
• Define expected output formats
• Avoid overly generic skills
• Keep SKILL.md concise (use references if needed)

• Treating skills as simple prompts
• Writing vague descriptions (skill won’t trigger)
• Overloading a single skill with multiple responsibilities
• Not defining output structure

• name (required)
• description (required)
• allowed-tools (optional)
• model (optional)

• Keep core logic in SKILL.md
• Move large content to:
  • references/
  • scripts/
  • assets/

Skills transform how we build with AI:

• From prompts → reusable systems
• From manual instructions → automatic behavior
• From experimentation → production-ready workflows

They are a key building block for:

• AI agents
• Automation pipelines
• Scalable AI systems

Curated links, documentation, and example repositories:

👉 Explore resources

Skills shift the mindset from:

“What should I ask the model?”

To:

“What capability should this system have?”

Contributions are welcome!

Ways to contribute:

• Add new skills or use cases
• Improve documentation and examples
• Report issues or suggest improvements

Feel free to open an issue or submit a pull request 🚀

I share practical insights about AI, agents, and real-world applications:

• LinkedIn (Profile): https://www.linkedin.com/in/elisa-terumi
• LinkedIn (Page): https://www.linkedin.com/company/exploring-artificial-intelligence
• Newsletter: https://exploringartificialintelligence.substack.com/
• Medium: https://medium.com/@elisa-terumi