Claude Skills are folders of instructions, scripts and resources that Claude loads only when a task needs them, such as Anthropic's built-in Excel, Word, PowerPoint and PDF skills or a custom skill that holds your brand rules. To use them on claude.ai, turn on Create and edit files in Settings, Capabilities, then manage skills from the Plugins page.
Anthropic's help center says Skills are on Free, Pro, Max, Ultra, Team and Enterprise plans, and they also work in Claude Code and the Claude API. Below: what a skill is, how to turn Skills on, add or upload one, how they differ from Projects, and how to stay safe. Screenshots we took on claude.ai on November 19, 2026, where Customize now opens a Plugins page the help articles do not show yet.
By Adi Leviim, Co-Founder at Infi Developments. 7+ years building AI productivity tools. Last reviewed November 2026.
What Are Claude Skills?
Anthropic defines Skills as "folders of instructions, scripts, and resources that Claude loads dynamically to improve performance on specialized tasks." A skill teaches Claude to do one job the same way every time, such as a report in your house style.
Skills use what Anthropic calls progressive disclosure: Claude uses each skill's description to decide which skills are relevant to your request and loads only the information it needs. You do not call a skill by name: "Claude determines when each skill is needed based on your request." There are three kinds:
Anthropic skills: built in for all users: Excel, Word, PowerPoint and PDF.
Partner skills: a directory with skills from Notion, Figma, Atlassian and others.
Custom skills: ones you or your organization write.
Skills follow the Agent Skills open standard (agentskills.io) and also work in the Claude for Excel, PowerPoint, Word and Outlook add-ins.
How to Turn On Skills in Claude
Skills need Claude's file and code tools, so start with a setting. On Free, Pro, Max and Ultra:
Open Settings, then Capabilities.
Switch on Create and edit files. Its label ends: "Skills need this on." The help article's older name is "Code execution and file creation".
Open your skills list and switch individual skills on or off.
On Team and Enterprise, an owner turns on "Cloud code execution and file creation" and "Skills" in Organization settings, Plugins & skills, Policy (on by default on Team).
How to Add Skills: Anthropic's, the Directory or Your Own
On claude.ai, Customize now opens the Plugins page, where Skills is one of three filters. In our November 19, 2026 screenshots:
A search box and the filters All, Connectors and Skills.
Your plugins lists skills you made, labelled "Skill, By you". Most used mixes connectors such as Slack and Notion with skill plugins such as frontend-design.
The Add menu offers Upload plugin, Add custom connector, Create a plugin, Create with Claude and Add marketplace.
Anthropic's help articles still describe the older layout: Customize, Skills, then "+" and "Browse skills" for the directory, or "+", "Create skill" and "Upload a skill" for your own ZIP file. The easiest way to build a skill is to ask Claude: Anthropic's skill-creator skill asks about your workflow, then creates and packages the files. On Pro, Max, Ultra and Team, Cowork in Claude for Mac can also record a skill while you show it a task (about 10 minutes); see what Claude Cowork is and how to use it.
Reuse a prompt, not a whole workflow? AI Toolbox (formerly Claude Toolbox) saves prompts you insert on claude.ai by typing // (10 on the free plan, unlimited on Premium) and runs prompt chains of up to 25 steps (2 chains free, unlimited on Premium). 40,000+ users on the Chrome Web Store.
A custom skill is a folder with a SKILL.md file that starts with YAML frontmatter holding a name and a description, zipped with the folder at the root. A minimal example:
brand-guidelines/
SKILL.md
resources/
---
name: brand-guidelines
description: Apply our brand colors, fonts and tone when creating documents or slides.
---
Steps Claude should follow, written as plain instructions.
Name the folder exactly like the skill's name: lowercase letters, numbers and hyphens, up to 64 characters, without "anthropic" or "claude".
Write a clear description. Claude uses it to decide when to load the skill; keep it under 200 characters for claude.ai.
Zip the folder itself, so the ZIP contains brand-guidelines/ at its root, not a subfolder.
Upload the ZIP as described above, then switch it on.
The help article writes skill.md; Anthropic's developer docs and Claude Code use SKILL.md. Uploads fail if the ZIP is too large, the folder and skill names differ, SKILL.md is missing, or the name or description has invalid characters. Skills cannot reference each other, but Claude can use several together. Custom skills are private until shared; on Team and Enterprise, share one from its three-dot menu (view-only for recipients) or Publish to org.
Skills vs Projects vs Custom Instructions
Projects and custom instructions are always loaded; skills load only when a task calls for them.
Skills you enable on claude.ai also load in Claude Code when you sign in with the same Claude account, on version 2.1.273 or later. Run /skills to see them under claude.ai sync. The sync is one-way and skipped when Claude Code uses an API key or a cloud provider.
Local skills: put a SKILL.md in ~/.claude/skills/<skill-name>/ for all your projects, or .claude/skills/<skill-name>/ for one repository.
Invoking: Claude uses skills when relevant, or you type /skill-name. Custom commands have been merged into skills.
Example skills:/plugin marketplace add anthropics/skills, then install document-skills@anthropic-agent-skills.
API: reference pptx, xlsx, docx or pdf by skill_id, or upload your own through /v1/skills. It needs the code execution tool; limits are 20 skills per request and 30 MB per skill.
Skills can run code, so Anthropic's advice is blunt: install skills only from trusted sources. It names the two main risks as "prompt injection, which allows Claude to be manipulated to execute unintended actions, and data exfiltration".
Read before enabling. For a skill from a less trusted source, "including one shared by a colleague", read every bundled file first.
No secrets inside. "Don't hardcode sensitive information (API keys, passwords)."
Network access differs: full, partial or none on claude.ai depending on settings, none on the API, full in Claude Code.
Enterprise organizations can turn on skill scanning for malicious content.
Compliance: "Agent Skills is not covered by ZDR arrangements."
Limitations
Anthropic's pages disagree on details: the help center includes the Free plan and caps descriptions at 200 characters, while the API docs list only paid plans for custom skills on claude.ai and allow 1,024 characters. No claude.ai upload size limit is published (the 30 MB figure is for the API). Labels may differ on your account. For the files Skills help Claude make, see our guides to Claude artifacts and uploading files to Claude. AI Toolbox does not create, install or run skills.
Frequently Asked Questions
What are Claude Skills in simple terms?
A Claude skill is a folder of instructions, and sometimes scripts or templates, that teaches Claude to do one task the same way every time. Claude uses each skill's description to decide when it is relevant and loads it only then. Anthropic ships Excel, Word, PowerPoint and PDF skills.
Are Claude Skills free?
Anthropic's help center lists Skills on Free, Pro, Max, Ultra, Team and Enterprise plans, and says Free users can enable example skills and upload their own. Skills need the Create and edit files capability switched on. Some older Anthropic pages still describe Skills as a paid-plan feature.
How do I turn on Skills in Claude?
Open Settings, Capabilities and switch on Create and edit files, which Claude labels as required for Skills. Then open your skills list, on the Plugins page in the current claude.ai layout, and switch individual skills on or off. On Team and Enterprise, an owner must also allow Skills in organization settings.
How do I create my own Claude skill?
Make a folder named after the skill with a SKILL.md file that starts with a name and short description in YAML frontmatter, zip the folder and upload it. You can also ask Claude to build one with Anthropic's skill-creator skill, which writes and packages the files for you.
What is the difference between Claude Skills and Projects?
Projects hold background files and instructions that Claude uses in every chat inside that project. Skills hold procedures that Claude loads only when a request needs them, in any chat. Use a project for knowledge about one topic, and a skill for a repeatable task such as formatting a report.
Are Claude Skills safe to install?
Only from sources you trust. Anthropic warns that a malicious skill can use prompt injection or leak data, and advises reading every file in a skill before enabling it, even one from a colleague. Never put API keys or passwords in a skill; Claude Code skills have full network access.
Bottom Line
Claude Skills package a repeatable task into a folder Claude loads only when needed. Turn on Create and edit files, let Anthropic's document skills work on their own, and add your own skill as a zipped SKILL.md folder or build one with Claude. Read every third-party skill before you switch it on.
A Full Stack Developer with 7+ years of experience building AI productivity tools. Leads product development and frontend architecture for AI Toolbox, the Chrome extension suite (ChatGPT, Gemini, Grok, and Claude modules) that helps users search, organize, and export their AI conversations.
Liked this guide? Get the next one first.
Join 40,000+ readers getting practical prompts, new features, and subscriber-only deals. No spam, unsubscribe in one click.
New features and product updates before anyone else
Time-saving prompts and workflows for ChatGPT, Gemini, Claude, and Grok