Claude Skills Explained: Best Skills & How to Use Them (2026)

1. What Are Claude Skills? Definition and Overview

A Claude Skill is a discrete, installable capability module that extends what Claude does by default. Instead of re-pasting the same instructions into every chat, you package them once and Claude applies them automatically when relevant. A skill is a folder of instructions you build once that Claude loads automatically when the task calls for it, so your preferences, workflows, and domain expertise are embedded in the skill, not re-pasted into every chat.

This is different from standard prompting or custom instructions. A one-off prompt handles a single request. A Skill is a reusable procedure Claude can reach for repeatedly across sessions and products.

Skills sit squarely within Anthropic's product roadmap for making Claude more capable and consistent. Anthropic unveiled Skills as a feature designed to let developers extend Claude with modular, reusable task components. They are available broadly across plan tiers and environments. Skills are available for users on Free, Pro, Max, Team, and Enterprise plans, this feature requires code execution to be enabled, and Skills are also available in beta for Claude Code users and for all API users using the code execution tool.

Claude Skills in One Sentence A Skill is a folder of instructions, and optional scripts and resources, that Claude loads automatically when your task needs it, giving consistent, expert-level output without re-explaining yourself every time.

2. How Claude Skills Work: The Three-Level Architecture

Skills work through a layered design that keeps Claude fast and focused. The key mechanism is progressive disclosure. Skills work through progressive disclosure, where Claude determines which skills are relevant and loads the information it needs to complete that task, helping to prevent context window overload.

Layer 1: The Skill Definition

The Skill definition is the content Claude reads. A skill is a folder, and inside it lives a SKILL.md file (required) and optionally a scripts/ directory for executable code, a references/ directory for documentation Claude loads as needed, and an assets/ directory for templates and supporting files. The SKILL.md file carries a name and a description, and the description is what tells Claude when to reach for the Skill. The description must include both what the Skill does and when Claude should use it.

Layer 2: The Execution Environment

Skills do not run in a vacuum. They rely on a secure sandbox to execute any bundled code or tool calls. Skills require the Code Execution Tool, which provides the secure environment they need to run. Each Skill also operates within defined boundaries. According to Anthropic, Skills are defined by a schema describing their inputs, outputs, and permissions. This is where scoping and permissions keep a Skill from doing more than it should.

Layer 3: The User or Org Context

The third layer is context: whether a Skill is personal to you or provisioned across your organisation. Custom Skills are individual to each user, they are not shared organization-wide, and cannot be centrally managed by admins. Org-provisioned Skills, by contrast, are pushed to everyone in a workspace, which is how teams personalise Claude at scale without compromising security controls.

Building AI into your workflow but tired of a 2-3% fee on every USD renewal? Bleap pays a flat 20% cashback on Claude, ChatGPT, and Gemini subscriptions and charges 0% FX fees, so your monthly AI spend works harder. Get the Bleap card →

3. Types of Claude Skills

Not all Skills are created equal. Four distinct categories exist, each with a different builder and governance model.

Anthropic-Built Skills

These are first-party Skills shipped and maintained by Anthropic. These are skills created and maintained by Anthropic, such as enhanced document creation for Excel, Word, PowerPoint, and PDF files, available to all users, and Claude invokes them automatically when relevant. Because Anthropic builds them, they carry the strongest quality and safety guarantees.

Partner Skills

Partner Skills are published by vetted third-party companies through Anthropic's Skills Directory. The Skills Directory features professionally-built skills from partners like Notion, Figma, Atlassian, and others, designed to work seamlessly with their respective MCP connectors, enabling powerful integrated workflows. Categories span productivity suites, design tools, and data platforms, and each partner Skill goes through a vetting process before listing.

Custom Claude Skills

Custom Skills are the ones you or your developers build for personal or team workflows. These are skills you or your organization create for specialized workflows and domain-specific tasks. Build a custom Skill when no existing option fits your exact workflow, and install an existing one when it already does. We expand on the build process in Section 10.

Org-Provisioned Skills

Enterprise admins can push Skills to every workspace member. Org-provisioned Skills exist to standardise how teams work. They distribute approved workflows consistently across all employees, ensure teams use standardized procedures and best practices, and deploy new capabilities without requiring individual uploads. Typical use cases include compliance bots, internal knowledge bases, and ticketing integrations.

4. Key Benefits of Using Claude Skills

• Extend Claude without retraining or fine-tuning. Skills add capability through instructions, not model changes, so you get new behaviour immediately.
• Automate repetitive, multi-step workflows in one conversation. A developer could create a Skill to fetch structured data from a company database, compose personalized email responses using CRM data, summarize meeting transcripts in a specific format, or trigger actions in third-party applications such as Slack or Notion.
• Connect live external data. Skills can call APIs, databases, and SaaS tools inside a single turn.
• Consistent behaviour across teams. Org-provisioned Skills mean everyone runs the same standardised procedure.
• Faster time-to-value. Skills improve Claude's consistency, speed, and performance on many tasks.
• Composable and portable. Multiple Skills can be stacked and used together, Claude will automatically identify which Skills it needs for a complex task and coordinate their use, and Skills use a unified format so you can write a Skill once and use it in the Claude Desktop app, Claude Code, and via the API.

5. Where Claude Skills Work: Platforms and Environments

Claude.ai (Web & Mobile)

On Claude.ai, Skills live in your settings and Claude invokes them automatically when relevant. You upload your own Skills as zip files through Settings and Features, available on Pro, Max, Team, and Enterprise plans with code execution enabled. Free-tier users get Anthropic's built-in Skills, but full custom-Skill uploads and org features unlock on paid plans.

Claude Code

Claude Code is Anthropic's agentic coding environment, and Skills fit naturally into a developer's workflow. Claude Code skills let you drop a SKILL.md into a project and have it apply every time you work in that repo, which is its own kind of relief if you have ever re-explained your commit format for the hundredth time. Examples include linting integrations, repo-aware code generation, and CI/CD hooks. Skills also follow an open standard, so your Claude Code extensions travel well. Skills are an open standard, and the Agent Skills specification is adopted by Claude Code, OpenAI Codex CLI, Cursor, Gemini CLI, and GitHub Copilot, so a skill you write or install works across all of these tools without modification.

Claude API

For custom application builders, Skills pass through the API alongside Anthropic's tool-use features. The Skills feature spans Claude's web app, the Claude Code environment, and the API, giving organizations the ability to author custom Skills through a new /v1/skills endpoint, manage versions in the console, and integrate them into their workflows. Once defined, Skills invoke dynamically. Once defined, they can be invoked dynamically through Claude's API, allowing seamless integration between the model and external systems. Keep rate limits and token cost in mind when chaining several Skills per request.

Building AI into your workflow but tired of a 2-3% fee on every USD renewal? Bleap pays a flat 20% cashback on Claude, ChatGPT, and Gemini subscriptions and charges 0% FX fees, so your monthly AI spend works harder. Get the Bleap card →

6. How to Find the Best Claude Skills: Directories and Categories

The official starting point is Anthropic's own repository. Anthropic published the official skills repository at github.com/anthropics/skills as a working reference for how skills should be structured. Its adoption has been striking. As of May 2026, the repo has 141,000+ stars and 16,000+ forks, making it one of the most-watched AI tooling repositories on GitHub.

Beyond the official directory, community-maintained collections list hundreds of Skills across engineering, marketing, product, compliance, and research categories, browsable by domain. Before you install anything from a less-trusted source, evaluate it carefully. Always read the full SKILL.md and audit every bundled script before installing community skills from less-trusted sources, and treat the anthropics/skills repository and partner skills from vetted publishers as safe defaults.

Best Claude Skills to try in 2026 (shortlist):

• Web search & real-time data, pull live information into a task
• Code interpreter / execution, run and test code in the sandbox
• PDF & document reader, parse and summarise files
• GitHub integration, sync issues, PRs, and repo context
• Calendar & scheduling assistant, automate planning tasks
• SQL / database query runner, turn questions into queries

One honest note on scope: as of April 8, 2026, Anthropic's Claude Code docs list five bundled skills, and Anthropic's official anthropics/skills marketplace currently exposes document-skills, example-skills, and claude-api. The ecosystem is real, but most people do not need dozens of Skills on day one.

7. How to Install and Activate Claude Skills: Step-by-Step

Installing Skills on Claude.ai

1. Open your account settings and navigate to the Skills area under Settings and Features.
2. Browse or search for the Skill you want.
3. Review the permissions and data access it requests.
4. Upload the Skill as a zip file, or install a listed one, completing any OAuth or API-key flow.
5. Activate the Skill in a new or existing conversation.
6. Confirm it is live with a quick test prompt.

Claude Skill Installation via API

1. Define the Skill or tool as a JSON object with its schema.
2. Pass the definition in the tools parameter of your API request.
3. Handle tool_use blocks in the assistant's response.
4. Return tool results and continue the conversation loop.

# Minimal Python sketch
import anthropic

client = anthropic.Anthropic()
response = client.messages.create(
model="claude-sonnet-4-5",
max_tokens=1024,
tools=[{"name": "my_skill", "description": "What it does and when to use it", "input_schema": {...}}],
messages=[{"role": "user", "content": "Run my skill on this data"}],
)

Admin Installation of Org-Provisioned Skills

Enterprise admins deploy Skills workspace-wide from the Admin Console. Select or upload the Skills to provision, set user permission levels and override policies, and push them to all members so nobody installs anything individually.

For Claude Code specifically, installation is even quicker. Claude asks you to choose a scope, where User installs it for all your projects and Project installs it for the current repo only, and that is it, the skill is now active and you can invoke it with /skill-name or let Claude load it automatically when relevant.

8. Top Claude Code Skills Worth Using

Claude Code deserves its own section because developers push Skills the hardest. You can install at three levels. You can install skills at three levels: enterprise (org-wide), personal (~/.claude/skills/), or project-specific (.claude/skills/).

• Shell Commander, run terminal commands directly from a conversation. Install at project scope for repo-specific tasks.
• GitHub Copilot Bridge, sync PR reviews and issue comments into your Claude Code session.
• Linting & Static Analysis, real-time code quality feedback as Claude writes.
• Test Runner, execute test suites and surface failures inline before you commit.
• Dependency Auditor, scan package.json or requirements.txt for known vulnerabilities.
• Dockerfile Generator, context-aware container configuration.

Practical tip: Skills are composable, so you can stack several in one session and Claude will coordinate which ones it needs. Start with bundled Skills first and add extras only when a real workflow demands them.

9. Claude Skills vs. Other Claude Features

New users mix these up constantly. Here is the clean version.

Skills vs. Claude MCP Tools

MCP (Model Context Protocol) is the low-level connectivity layer; Skills are the user-facing abstraction that often sits on top. Anthropic's own analogy is helpful. MCP provides the professional kitchen, meaning access to tools, ingredients, and equipment, while Skills provide the recipes and step-by-step instructions for creating something valuable. Use MCP tools directly when you need raw tool access, and wrap them in a Skill when you want a repeatable procedure.

Skills vs. Projects

Projects store context and files; Skills add active, tool-based capability. They are not rivals. They aren't competitors, they sit at different layers, and a real workflow usually uses all three. A Project holds the material, and a Skill acts on it.

Skills vs. Custom Instructions

Custom Instructions shape Claude's tone and personality; Skills add tool-based actions. Run both at once: instructions set the voice, Skills do the work.

10. How to Create Your Own Custom Claude Skills

Prerequisites

You need an Anthropic account with API access, a basic grasp of JSON tool definitions (or the visual builder), and any API credentials for the integration you are targeting.

Designing Your Skill

Keep it tight: one Skill, one job. Map the trigger conditions and expected inputs and outputs, then write the manifest with a clear name, a strong description, and a parameters schema. The description matters most, because it is what tells Claude when to use the Skill. Get started fast: the simplest way to create your first skill is to create a directory, add a SKILL.md file with YAML frontmatter, and Claude discovers it automatically.

Building and Testing

Iterate in the Anthropic Workbench or an API sandbox. Write a system prompt that activates the Skill correctly, then test the edge cases: unexpected inputs, API errors, and rate limits. Anthropic's structure is designed to make this quick. By the end, you will be able to build a working skill in one sitting, which is exactly what Anthropic's official guide promises for anyone who follows the structure correctly.

Publishing and Sharing

Publish partner-grade Skills to the Skills Directory via the partner programme, share custom Skills within your organisation, or open-source them on GitHub for the community. Because Skills are open by nature, transparency is the norm. Skills are not models, plugins in the WordPress sense, or paid add-ons; they are open-source markdown instructions plus supporting files, and you can read every one of them on GitHub before you install anything.

11. Security Considerations and Limitations

Permission Scoping and Least-Privilege Design

Only request the permissions a Skill genuinely needs. Because each Skill runs within a defined schema of inputs, outputs, and permissions, least-privilege design is built into the format. Store API keys securely and keep OAuth scopes minimal.

Data Privacy and Retention

Review what data a Skill can access and for how long. Enterprise teams deploying org-provisioned Skills should account for data residency requirements before pushing a Skill workspace-wide.

Known Limitations

• Skills cannot override Anthropic's core safety policies.
• When a Skill is not relevant to a user's request, Claude will bypass it, ensuring that only pertinent instructions and resources are loaded, but chaining many Skills in one turn can still add latency.
• Large Skill payloads can eat into the context window.
• Availability differs by plan tier and requires code execution enabled.

Responsible Use Guidelines

Skill builders must comply with Anthropic's usage policy. If you encounter a problematic partner Skill, use the reporting channels to flag it, and always audit third-party scripts before installing.

12. Claude Skills Resources and Community

• Official documentation, the Claude Platform docs and Help Center Skills pages.
• Community directories, the official anthropics/skills repository plus community collections on GitHub.
• Developer forums and Discord, where Skill builders share patterns and troubleshoot.
• Newsletters and blogs, tracking new Skill and partner releases.

Recommended learning path for developers new to Claude AI skills:

1. Read the API tool-use docs.
2. Build a simple single-action Skill.
3. Explore existing open-source Skills for patterns.
4. Graduate to multi-step, org-provisioned Skill deployment.

Running Claude, ChatGPT, and Gemini side by side every month? Those USD subscriptions add up, and most cards tack on 2-3% FX on every renewal. Bleap charges 0% FX fees and pays a flat 20% cashback on all three. Get the Bleap card →

Frequently Asked Questions About Claude Skills

What is the difference between Claude Skills and Claude MCP tools?

Skills are user-installable capability modules made of instructions and optional scripts, while MCP is the low-level protocol that powers tool connectivity. In Anthropic's own framing, MCP is the kitchen and Skills are the recipes. Skills often wrap MCP tools to make them easier to use.

How do I install a custom Claude skill on Claude.ai?

Open Settings and Features, upload your Skill as a zip file, review the permissions it requests, complete any OAuth or API-key flow, then activate it in a conversation and confirm with a test prompt. See Section 7 for the full walkthrough.

Are Claude Skills available on the free plan?

Skills are available for users on Free, Pro, Max, Team, and Enterprise plans, and this feature requires code execution to be enabled. That said, custom Skill uploads are available on Pro, Max, Team, and Enterprise plans with code execution enabled, so free-tier users mainly get Anthropic's built-in Skills.

Can I build and sell my own Claude Skills?

Yes. You can open-source Skills freely on GitHub, or publish professionally-built Skills through the partner programme to appear in the Skills Directory. Partner listings go through a vetting process before they go live.

What are org-provisioned skills and who manages them?

They are admin-deployed Skills pushed to all workspace users, managed from the Admin Console. They distribute approved workflows consistently across all employees, ensure teams use standardized procedures, and deploy new capabilities without requiring individual uploads.

How do Claude Code skills differ from regular Claude Skills?

Claude Code Skills use the same SKILL.md format but are optimised for agentic coding: shell access, repo context, and test running. Claude Code skills let you drop a SKILL.md into a project and have it apply every time you work in that repo. They install via Claude Code's own plugin browser, the skills CLI, or a project config directory.

Conclusion: Unlocking the Full Power of Claude AI Skills

Claude Skills come in four flavours: Anthropic-built for safety-guaranteed basics, partner Skills for vetted integrations, custom Skills for your own workflows, and org-provisioned Skills for team-wide consistency. All four rest on the same three-level architecture: a Skill definition, a secure execution environment, and user or org context. For enterprise teams, org-provisioned Skills are the standout, turning one good workflow into a standard everyone follows.

Your next moves are simple: explore the official Skill directory today, try one Claude Code Skill in a live repo, and prototype a custom Skill with the API. The Claude Skills ecosystem is expanding fast, and staying current is a genuine competitive advantage.

One last practical tip. Whichever AI tools you rely on, pay smart. Those Claude, ChatGPT, and Gemini subscriptions are billed monthly in USD, and a typical card quietly adds 2-3% on every renewal. With Bleap you skip the FX fees entirely and earn a flat 20% cashback on Claude, ChatGPT, and Gemini, using a self-custodial Mastercard with no subscription of its own.

Stop paying FX fees on your AI stack. Bleap gives you 0% FX fees on USD subscriptions and a flat 20% cashback on Claude, ChatGPT, and Gemini, with a self-custodial Mastercard and no monthly fee. Get the Bleap card →