Introduction
Anthropic has announced Claude Skills, an innovative system that enables the Claude language model to acquire specialized capabilities for specific tasks. Skills are folders containing instructions, scripts, and resources that Claude loads only when needed, dramatically improving performance on targeted activities like managing Excel spreadsheets or applying brand guidelines.
Claude Skills represents a significant evolution in the approach to artificial intelligence: instead of loading all information upfront, the model dynamically accesses only the relevant skills for the task at hand, maintaining speed and efficiency.
What Are Claude Skills
Skills are packages of specialized knowledge that Claude can use to enhance its performance. While working, Claude scans available Skills to find relevant matches and, when it identifies a pertinent one, loads only the minimal information and files necessary.
Key Characteristics
- Composability: Skills stack together, with Claude automatically identifying which are needed and coordinating their use
- Portability: They use the same format everywhere, allowing you to build once and use across Claude apps, Claude Code, and API
- Efficiency: Only loads what's needed, when it's needed, keeping Claude fast
- Power: Can include executable code for tasks where traditional programming is more reliable than token generation
How Skills Work
A Skill is conceptually very simple: it's a Markdown file telling the model how to do something, optionally accompanied by extra documents and pre-written scripts that the model can run to help accomplish tasks.
At the start of a session, Claude's various harnesses scan all available skill files and read a short explanation for each one from the frontmatter YAML in the Markdown file. This approach is extremely token efficient: each Skill only takes up a few dozen extra tokens, with full details loaded only when the user requests a task that the Skill can help solve.
Technical Requirements
The Skills mechanism depends entirely on the model having access to a filesystem, tools to navigate it, and the ability to execute commands in that environment. This dependency is significant but unlocks a bewildering amount of new capability.
Availability Across All Claude Products
Claude Apps
Skills are available to Pro, Max, Team, and Enterprise users. Anthropic provides Skills for common tasks like document creation, customizable examples, and the ability to create custom Skills. Claude automatically invokes relevant Skills based on the task—no manual selection needed. Users can even see Skills in Claude's chain of thought as it works.
Creating Skills is simple thanks to the "skill-creator" skill that provides interactive guidance: Claude asks about your workflow, generates the folder structure, formats the SKILL.md file, and bundles necessary resources, without requiring manual file editing.
Claude Developer Platform (API)
Agent Skills can now be added to Messages API requests and the new /v1/skills endpoint gives developers programmatic control over custom skill versioning and management. Skills require the Code Execution Tool beta, which provides the secure environment they need to run.
Developers can use Anthropic-created Skills to have Claude read and generate professional Excel spreadsheets with formulas, PowerPoint presentations, Word documents, and fillable PDFs. Custom Skills can also be created to extend Claude's capabilities for specific use cases.
Claude Code
Skills extend Claude Code with your team's expertise and workflows. Install Skills via plugins from the anthropics/skills marketplace, with Claude loading them automatically when relevant. Skills can be shared through version control with your team, or manually installed by adding them to ~/.claude/skills.
Skills vs MCP: A Revolutionary Comparison
Model Context Protocol (MCP) has attracted enormous buzz since its initial release in November 2024. However, over time significant limitations have emerged, particularly regarding token usage: GitHub's official MCP alone consumes tens of thousands of context tokens, leaving precious little space for the LLM to do useful work once you've added a few more.
Advantages of Skills
Skills offer the same advantages as CLI tools: LLMs know how to call cli-tool --help, which means you don't have to spend many tokens describing how to use them. With Skills, you don't even need to implement a new CLI tool: just drop a Markdown file describing how to do a task, adding extra scripts only if they'll help make things more reliable or efficient.
"The core simplicity of the Skills design is why I'm so excited about it. MCP is a whole protocol specification covering hosts, clients, servers, resources, prompts, tools, sampling, roots, elicitation and three different transports. Skills are Markdown with a tiny bit of YAML metadata and some optional scripts."
Practical Applications and Use Cases
The potential applications of Skills are extremely broad. Considering just the data journalism domain, one can imagine a folder of Skills covering tasks like:
- Where to get US census data from and how to understand its structure
- How to load data from different formats into SQLite or DuckDB using appropriate Python libraries
- How to publish data online, as Parquet files in S3 or pushed as tables to Datasette Cloud
- A skill defined by an experienced data reporter on how to find interesting stories in a new dataset
- A skill describing how to build clean, readable data visualizations using D3
This amounts to building a "data journalism agent" with a folder full of Markdown files and maybe a couple of example Python scripts.
Claude Code as a General Agent
Claude Code, despite its name, is not purely a coding tool: it's a tool for general computer automation. Anything you can achieve by typing commands into a computer is something that can now be automated by Claude Code. Skills make this aspect much more obvious and explicit, positioning Claude Code as a true general agent.
Sharing and Future of Skills
One of the most exciting things about Skills is how easy they are to share. Many Skills will be implemented as a single file, while more sophisticated ones will be a folder with a few more files. Anthropic has already published Agent Skills documentation and a Claude Skills Cookbook.
Another interesting aspect of the Skills design is that there is nothing preventing them from being used with other models. You can grab a Skills folder right now, point Codex CLI or Gemini CLI at it and say "read pdf/SKILL.md and then create me a PDF describing this project" and it will work, despite those tools and models having no baked-in knowledge of the Skills system.
Future Prospects
Anthropic is working on simplified skill creation workflows and enterprise-wide deployment capabilities, making it easier for organizations to distribute Skills across teams. A Cambrian explosion in Skills is expected that will make this year's MCP rush look pedestrian by comparison.
Security Considerations
It's important to note that this feature gives Claude access to execute code. While powerful, it means being mindful about which Skills you use: stick to trusted sources to keep your data safe. The need to make safe coding environments available to LLMs is fundamental, with particular attention to sandboxing to limit damage from attacks such as prompt injections.
Conclusion
Claude Skills represents a paradigm shift in how we interact with large language models. Their conceptual simplicity, combined with practical power, makes them superior to more complex protocols like MCP. By outsourcing the hard parts to the LLM harness and the associated computer environment, Skills adopt a very sensible strategy that fully leverages what we've learned about LLMs' ability to run tools over the last couple of years.
With the ability to create, share, and use Skills across different platforms and even different models, Anthropic has introduced a system that could define the future of AI-based automation. The true strength of Skills lies in their simplicity: Markdown files with YAML metadata and optional scripts that let LLMs do what they do best.
FAQ
What are Anthropic's Claude Skills?
Claude Skills are folders containing Markdown instructions, scripts, and resources that Claude loads dynamically when needed to improve performance on specific tasks, like working with Excel or following brand guidelines.
How do you create custom Claude Skills?
You can create Skills using the "skill-creator" skill that provides interactive guidance through the process, generating folder structure and SKILL.md file, or manually by creating Markdown files with YAML metadata and optional scripts.
Do Claude Skills work with all Anthropic products?
Yes, Skills use the same format and work across Claude apps (Pro, Max, Team, Enterprise), Claude Developer Platform API, and Claude Code, ensuring complete portability.
What's the difference between Claude Skills and MCP?
Skills are simpler and more efficient: they use Markdown files with YAML metadata instead of a complex protocol, consume fewer tokens, and can work with other AI models without built-in knowledge of the system.
Do Claude Skills require a coding environment?
Yes, Skills depend on the availability of a filesystem, navigation tools, and the ability to execute commands, but this requirement unlocks an enormous amount of new capabilities for automation.
How does Anthropic ensure security with Skills?
Since Skills allow code execution, it's crucial to use only Skills from trusted sources. Anthropic is working on sandbox environments to limit damage from attacks like prompt injections while keeping data safe.