The most useful Claude Skills repositories on GitHub right now are Anthropic's own anthropics/skills, a handful of focused community collections (marketing, science, design, code-intelligence), and two curated lists that link to hundreds more. This article covers nine, with the install command from each README, the licence GitHub reports and the star count on 2026-11-19. Read the safety section before you install anything, because a Skill can bundle scripts that Claude will run.
Who this is for and what you will have at the end
This is for people who use Claude Code or claude.ai and want ready-made Skills instead of writing every SKILL.md from scratch. You will leave with a short list of repositories to try, the exact command to install each, and a checklist for deciding whether a repository is safe enough to install.
If you have not used Skills before, read What Are Claude Skills? How to Install and Use One Step by Step first. It explains the folder format and where Claude Code looks for Skills.
How Skills work, in two paragraphs
A Skill is a folder with a SKILL.md file. Anthropic's documentation says the frontmatter requires name and description; the name is limited to 64 characters of lowercase letters, numbers and hyphens, and the description to 1,024 characters. Claude loads only the name and description at startup, and reads the rest of the folder when your request matches the description. Anthropic puts the metadata cost at roughly 100 tokens per Skill.
In Claude Code, Skills are filesystem-based: put them in ~/.claude/skills/ for personal use or .claude/skills/ inside a project. They can also be shared as Claude Code plugins. In claude.ai, custom Skills are uploaded as zip files and need a Pro, Max, Ultra, Team or Enterprise plan with code execution enabled. Custom Skills do not sync across surfaces, so a Skill installed in Claude Code is not automatically in claude.ai.
Safety checklist before you install any repo
Anthropic's Skills documentation is direct: use Skills only from trusted sources, meaning ones you created or obtained from Anthropic, and audit thoroughly if you must use anything else. A malicious Skill can direct Claude to call tools or run code that does not match its stated purpose, and in Claude Code Skills have the same network access as any other program on your computer.
Run this check on every repository below, and on any you find yourself:
- Open
SKILL.mdand read it. Does the description match what the body tells Claude to do? - List every other file in the skill folder. Read each script. Look for network calls, reads of files outside the project, and anything that sends data elsewhere.
- Look for instructions that fetch content from external URLs at run time. Anthropic notes that fetched content can carry malicious instructions, and that even a trustworthy Skill can be compromised if its external dependencies change.
- Check the licence file, and the licence field in the skill itself where it exists.
- Check the date of the last commit and whether issues are being answered.
- Install one skill, not the whole collection. Remove it if it never triggers.
- Pin a release or commit where the README offers a way to do so, so an update cannot silently change what runs.
The nine repositories
Star counts and licences below come from the GitHub API on 2026-11-19. Stars measure popularity, not review. Install commands are quoted from each README as of the same date.
| Repository | Best for | Licence (GitHub API) | Stars on 2026-11-19 |
|---|---|---|---|
| anthropics/skills | Official examples and document skills | None reported at repo level; many skills Apache 2.0 per README | 197,294 |
| wshobson/agents | Plugins, agents and skills for software teams | MIT | 42,670 |
| coreyhaines31/marketingskills | Marketing, SEO, copywriting, CRO | MIT | 61,300 |
| K-Dense-AI/scientific-agent-skills | Scientific and research workflows | MIT | 57,740 |
| Nutlope/hallmark | Front-end design quality | MIT | 57,626 |
| Graphify-Labs/graphify | Turning a codebase into a knowledge graph | Apache-2.0 | 150,247 |
| sickn33/agentic-awesome-skills | Searchable catalog of thousands of skills | MIT (imported skills vary) | 49,970 |
| ComposioHQ/awesome-claude-skills | Curated list | None reported by API; README says Apache 2.0 | 79,831 |
| hesreallyhim/awesome-claude-code | Curated list of Claude Code resources | NOASSERTION (Other) | 58,321 |
1. anthropics/skills
What it does. Anthropic's public repository for Agent Skills. Its README describes skill examples in four groups: Creative and Design, Development and Technical (including testing web apps and MCP server generation), Enterprise and Communication, and Document Skills (docx, pdf, pptx, xlsx). It also holds the Agent Skills specification in a spec folder and a skill template in a template folder, which is the best starting point if you want to write your own.
Who it is for. Everyone. It is the one source Anthropic's documentation explicitly names as a trusted origin for Skills.
Install. From the README:
/plugin marketplace add anthropics/skills /plugin install document-skills@anthropic-agent-skills /plugin install example-skills@anthropic-agent-skills The README says the example skills are already available to paid plans in claude.ai.
Licence and caveats. The README says many skills are open source under Apache 2.0, including the document skills, which it shares as a reference for more complex skills. The GitHub API reports no repository-level licence, so check the skill folder you use. The README also states the skills are provided for demonstration and educational purposes and should be tested before you rely on them.
2. wshobson/agents
What it does. A catalog of plugins that bundle agents, skills and commands, written in Markdown. The README states 93 plugins in the catalog (91 local plus 2 external), 199 agents and 185 skills in the local source. Example plugin names include python-development, javascript-typescript, developer-essentials and security-scanning.
Who it is for. Developers who want a role-based toolkit for a language or task rather than one-off skills.
Install.
/plugin marketplace add wshobson/agents /plugin install python-development@claude-code-workflows Licence. MIT; the README says external plugins have their own licences. Install a single plugin first, because the catalog is large.
3. coreyhaines31/marketingskills
What it does. The README states 53 marketing skills, with examples such as cro, copywriting, ab-testing, seo-audit and product-marketing. The README describes product-marketing as the foundation the other skills check first, so install it alongside whatever you actually want.
Who it is for. Marketers, founders and small-business owners who use Claude for landing pages, copy and SEO reviews.
Install. The README lists nine routes. The ones relevant here:
npx skills add coreyhaines31/marketingskills npx skills add coreyhaines31/marketingskills --list npx skills add coreyhaines31/marketingskills --skill cro copywriting /plugin marketplace add coreyhaines31/marketingskills /plugin install marketing-skills For claude.ai, the README says to download a <name>-claude-web.zip from Releases and upload it under Customize, then Skills. MIT licensed.
4. K-Dense-AI/scientific-agent-skills
What it does. A library the README describes as 177 scientific and research skills. Each skill has a SKILL.md with its purpose, workflow and version metadata. The README is explicit that installing the skill files does not install their dependencies, so each skill's own setup instructions still apply.
Who it is for. Researchers, analysts and students working in lab, bioinformatics or data-heavy fields.
Install.
npx skills add K-Dense-AI/scientific-agent-skills gh skill install K-Dense-AI/scientific-agent-skills The gh skill install route requires GitHub CLI v2.90.0 or later according to the README, and supports pinning a release with --pin.
Licence and caveats. The repository is MIT, but each skill carries its own licence in the license field of its SKILL.md. The README itself warns: "Skills can execute code and influence your coding agent's behavior. Review what you install."
5. Nutlope/hallmark
What it does. A design skill aimed at avoiding generic AI-looking front-end output. According to the README it installs for Claude Code, Codex CLI, Cursor and OpenCode. Its optional browser checks need puppeteer-core and Chrome; without them the static checks still run.
Who it is for. Developers and designers who generate UI with Claude Code and want more distinctive results.
Install.
npx skills add nutlope/hallmark -g --agent claude-code The README says -g installs for every project and that Node 18 or later is required. A plugin route is also documented:
/plugin marketplace add nutlope/hallmark /plugin install hallmark@hallmark It is then invoked as /hallmark:hallmark. MIT licensed; the README notes that its photo library uses Wikimedia Commons images under their own licences.
6. Graphify-Labs/graphify
What it does. A Claude Code skill. You run /graphify, and per the README it reads your files, builds a knowledge graph and returns structure. It accepts code, PDFs, Markdown and images. Output goes to a graphify-out/ folder, including an interactive HTML graph, an Obsidian vault, a GRAPH_REPORT.md summary and a graph.json.
Who it is for. Developers joining an unfamiliar codebase, and anyone trying to see how a pile of documents connects.
Install. The README requires Claude Code and Python 3.10 or later:
pip install graphifyy && graphify install The README says the PyPI package is temporarily named graphifyy while the CLI and skill command stay graphify. It also refers to the project under a different owner name, safishamsi/graphify, so confirm you are on the repository you intend and verify the package source before running pip install.
Caveats. The README says it runs locally, but also that docs and images are processed through Claude, meaning that content is sent to Claude. Do not point it at files you are not allowed to send there. GitHub reports Apache-2.0.
7. sickn33/agentic-awesome-skills
What it does. A large library the README describes as 3,022+ installable SKILL.md playbooks, with a local catalog and a CLI. It states the core tool lets an agent search the catalog and record chosen skills, and that it does not rank or recommend skills.
Who it is for. People who want to browse a wide range before choosing, and are prepared to vet what they pick.
Install. The README's worked example uses the GitHub CLI against a single skill:
gh skill install sickn33/agentic-awesome-skills skills/brainstorming/SKILL.md --agent codex --scope user --pin v21.0.0 That example targets Codex. The README lists claude-code as the --agent value for Claude Code, but it shows no complete Claude Code example, so check its plugin guide.
Caveats. The README says some listed skills are paid or hosted services, and some call external MCP servers or APIs. It also says structural validity does not certify safety. Original code is MIT; imported skills keep their own licences.
8. ComposioHQ/awesome-claude-skills
What it does. A curated list of Skills and plugins, organized by category: Document Processing, Development and Code Tools, Data and Analysis, Business and Marketing, Communication and Writing, Creative and Media, Productivity and Organization, Collaboration and Project Management, Security and Systems, and App Automation via Composio. The README says it lists over 1,000 skills and plugins.
Who it is for. Anyone looking for a Skill in a specific area who wants a starting map.
Install. There is no single install; each entry links elsewhere. The README's manual Claude Code route is:
mkdir -p ~/.config/claude-code/skills/ cp -r skill-name ~/.config/claude-code/skills/ That path differs from the ~/.claude/skills/ location in Anthropic's documentation, so follow Anthropic's path unless you have confirmed otherwise. The README also warns individual skills may carry different licences. The API reported no licence while the README mentions Apache 2.0, so treat the licence as unconfirmed and check each skill.
Caveats. The App Automation section is built around Composio's own tooling, which is a vendor product; read what it connects to before you enable it.
9. hesreallyhim/awesome-claude-code
What it does. A hand-picked list of Claude Code resources. Its sections include Skills, Agent Orchestration, Status Lines, Memory and Context Persistence, Security, Testing and Design and UI/UX, among others.
Who it is for. Claude Code users who want more than Skills: hooks, status lines, workflows and tooling.
Install. None; it is a list. The README does not describe a formal vetting process. It states an emphasis on code quality, security and originality, and many entries comment on their own security or testing. One entry carries an explicit note that its scripts contact Baidu and Yandex, which shows why reading the scripts matters.
Licence. GitHub reports NOASSERTION ("Other"). The README is very large, so this summary covers its opening sections only.
Choosing where to start
| If you want... | Start with |
|---|---|
| The safest, best-documented examples | anthropics/skills |
| A role-based toolkit for software work | wshobson/agents |
| Marketing and SEO workflows | coreyhaines31/marketingskills |
| Research and lab workflows | K-Dense-AI/scientific-agent-skills |
| Better front-end output from Claude Code | Nutlope/hallmark |
| A map of an unfamiliar codebase | Graphify-Labs/graphify |
| To browse widely before deciding | the two awesome lists |
Installing safely, step by step
- Pick one skill and open its folder on GitHub.
- Read
SKILL.md, then every script and reference file. - Check the licence and the last commit date.
- Install it for a single project first (
.claude/skills/or a project-scoped install) rather than for your whole machine. - Ask Claude a question that should trigger it. If the description is vague the Skill may never load; that is a quality signal.
- Review any command Claude proposes to run from the Skill before approving it.
- Remove Skills you do not use. Each installed Skill adds its name and description to Claude's context.
Troubleshooting
The Skill never triggers
Claude matches your request against the description. A vague description is the most common cause. Rephrase your request to include words from the description, or rewrite the description in a local copy so it says what the Skill does and when to use it.
A new skills folder is not picked up
Claude Code picks up edits to existing skill folders, but a brand-new top-level skills directory may need a reload. See the Claude Code Skills documentation linked below for the current command.
The Skill works in Claude Code but not in claude.ai
Anthropic documents that custom Skills do not sync across surfaces. Upload a zip separately in claude.ai, and make sure code execution is enabled.
The plugin command is not recognized
Plugin commands require a Claude Code version that supports plugins. Update Claude Code and run the command from the README exactly as written.
What we checked
- Queried the GitHub API on 2026-11-19 for
anthropics/skills. It returned 197,294 stars, anulllicence field and a last push of 2026-11-16. - Ran a GitHub API search for "claude skills" sorted by stars on 2026-11-19 and took stars, licence and last-push dates for the other repositories from that result.
- Opened each README on github.com on 2026-11-19 and copied the install commands, counts and licence statements quoted above.
- Cross-checked Skills file format,
nameanddescriptionlimits, the~/.claude/skills/and.claude/skills/paths and the security guidance against Anthropic's Agent Skills overview page on platform.claude.com on 2026-11-19. - Found two licence discrepancies between README and API (ComposioHQ/awesome-claude-skills and, at repo level, anthropics/skills) and reported both as unresolved.
- Noted that the Graphify README refers to a different owner name from the repository we queried; we did not investigate the history.
- Not checked: we did not install or run any Skill, did not read every script in these repositories and did not open the full awesome-claude-code README, so none of this is a security review.
Last tested: 2026-11-19
References
- Anthropic, Agent Skills overview: https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview
- Claude Code, Use Skills: https://code.claude.com/docs/en/skills
- https://github.com/anthropics/skills
- https://github.com/wshobson/agents
- https://github.com/coreyhaines31/marketingskills
- https://github.com/K-Dense-AI/scientific-agent-skills
- https://github.com/Nutlope/hallmark
- https://github.com/Graphify-Labs/graphify
- https://github.com/sickn33/agentic-awesome-skills
- https://github.com/ComposioHQ/awesome-claude-skills
- https://github.com/hesreallyhim/awesome-claude-code
This guide was researched and drafted with AI assistance and checked against the vendor documentation listed in its References. Read the maintained version and related guides on EpiFive.


