2026 guide
Claude Skills Marketplace: Where to Find Claude Skills
There is no single official “Claude skills marketplace.” A skill is just a folder with a SKILL.md file, and you get one by copying it into place. This guide maps the surfaces that actually exist — Anthropic's own, and the community directories — then covers how installation works and what to check before you add a skill.
Quick answer
To find Claude skills, start with Anthropic's own surfaces — claude.com/skills for the format and claude.com/marketplace for plugins and connectors. For a plain SKILL.md skill there is no package manager: you copy the folder into ~/.claude/skills/ (personal) or a repo's .claude/skills/ (project). The rest of the “marketplaces” you see in search results are independent community directories — useful for discovery, but not run or endorsed by Anthropic, and a listing is not a security review.
What a Claude skill actually is
A Claude skill is a folder that contains a SKILL.md file. Claude adds that folder to its toolkit and loads it when the skill's description matches the task, or you invoke it directly as /skill-name. There is no binary and no server — the artifact is a Markdown file the model reads. (Source: code.claude.com/docs/en/skills.)
Custom commands and skills are the same mechanism. Both .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md create a /deploy command and behave the same way. If you already have commands, you already have skills.
One format, two tools: the open standard
Claude Code skills follow the Agent Skills Open Standard (agentskills.io), which per Anthropic's docs is designed to work across multiple AI tools. That is the reason a skill is worth investing attention in: the folder you build is not locked to one product.
It is not 100% portable, though, and it is worth being precise about where the seams are. Claude Code adds extensions on top of the standard — invocation control, subagent execution, and dynamic context injection. Those frontmatter fields and behaviours are Claude Code-specific. A skill that leans on them will run in Claude Code but needs those assumptions removed before another runtime can use it.
The practical split: the workflow (what the skill tells the agent to do) is portable; the packaging and invocation controls are not.
Where to find Claude skills today
There are two kinds of sources, and mixing them up is the most common source of confusion. First-party surfaces explain the format and host plugins and connectors. Everything else is a community directory that indexes public skills.
Anthropic's own surfaces
claude.com/skills
Anthropic's own explainer for agent skills: what a skill is, a skill-creator, and a worked SKILL.md example. Start here to understand the format.
claude.com/marketplace/plugins
Anthropic's plugin marketplace. As of 13 Oct 2026 it lists 342 plugins, with “Anthropic verified” badges on first-party entries. A plugin is a packaged bundle, not a bare skill.
claude.com/marketplace
The broader Anthropic marketplace: connectors and plugins (2,000+ listed), partner agents and products, and service partners.
Community directories
Cult of Claude (cultofclaude.com)
Run by FDF Labs and independent of Anthropic. It crawls public GitHub every few hours; on 13 Oct 2026 it listed ~35,133 skills and 19,255 agents. Free to browse, and every entry links back to its source repo, which is the useful part — you can read the code before you trust the listing.
Claude Market (claudemarket.ai)
A directory of 23,000+ entries covering skills, MCP servers, plugins, and connectors. It ships an `npx claudemarket search` CLI and states it has 27,000+ monthly visitors.
AugmentClaude (augmentclaude.com)
Curated bundles rather than a raw index — someone has already grouped skills by job, which trades breadth for a shorter list.
Other community indexes
claude-marketplace.com, skillsclaude.org, SkillsMP (skillsmp.com), mcpservers.org, buildwithclaude.com, awesomeskill.ai, and agent37.com all list skills, MCP servers, or both. They differ mainly in how they are seeded and how often they refresh, so treat each as a starting point, not a source of truth.
To be explicit: Anthropic does not run or endorse these community directories. Being listed is not a security review, and none of these indexes vouches for correctness. They are discovery tools — the source repo they link to is where your judgment starts.
How installation actually works
For a plain skill there is no install command. You place a folder in the right scope and Claude picks it up. Skills synced from claude.ai land in a separate synced subfolder under ~/.claude/skills/, which is worth knowing so you do not edit a file that a sync will overwrite.
Plain skill — personal
Copy the folder into ~/.claude/skills/Plain skill — project
Keep it in the repo at .claude/skills/ so it travels with the project.Subagent
A single Markdown file in ~/.claude/agents/Plugin
Added through a plugin marketplace rather than by copying files.
What to check before you install a skill
A skill is instructions the agent will follow, and it can carry more than text. The rule one directory operator puts on their own site is the one to adopt everywhere:
Read the file before you run the command: a skill can pre-approve tools and attach hooks.
- Open the SKILL.md and any scripts or references it points to.
- Check whether it pre-approves tools or attaches hooks.
- Follow the entry back to its source repo and check the last commit date.
- Confirm the skill's description matches the job you actually want done.
Read the SKILL.md before you run anything
The listing is marketing; the file is the truth. Open the SKILL.md (and any references or scripts it points to) and read what it actually instructs the agent to do.
Look for pre-approved tools and hooks
A skill can pre-approve tool calls and attach hooks that run on its behalf. Treat that as code review, not content reading.
Confirm the source repo
Directories like Cult of Claude link to the upstream repo. Follow that link and check whether the skill is actively maintained before you depend on it.
Remember a listing is not a security review
No directory vouches for correctness. Cult of Claude states plainly that its score “rewards documentation and adoption, not correctness, and it is never a security review” — apply the same skepticism to every other index.
Run the same SKILL.md in Hermes Agent
Hermes Agent skills are also SKILL.md-based, so the file you found in a Claude directory is not a dead end if you work in a different runtime. The migration is usually small: keep the workflow, strip the Claude Code-specific assumptions — invocation control, subagent execution, dynamic context injection — and test it on a real task before you rely on it.
The step-by-step procedure lives in the migration guide; the workflow for choosing and auditing a skill is in the skills guide.
Related Hermes Agent guides
Hermes Agent skill marketplace
How skills are discovered and installed in Hermes Agent.
Hermes skill library
Browse reusable SKILL.md-based procedures for the Hermes runtime.
Skills Hub: search, inspect, install, audit
The correct workflow for adopting a skill — search, inspect, test, then keep.
Hermes Agent vs Claude Code
How a broad agent runtime compares with a coding-focused Claude tool.
How to create a skill
Write your own SKILL.md so it loads reliably and stays auditable.
FAQ
Is there an official Claude skills marketplace?
Anthropic runs claude.com/skills (what skills are, plus a skill-creator) and claude.com/marketplace (connectors and plugins, 2,000+ listed). For a bare SKILL.md skill there is no package manager or central store — you copy the folder into ~/.claude/skills/ or a repo's .claude/skills/. Most “skills marketplace” results you find in search are community directories, not Anthropic products.
Are community Claude skills marketplaces safe?
Treat them like any other source of executable instructions. Anthropic does not run or endorse the community directories, and being listed is not a security review. A skill can pre-approve tools and attach hooks, so read the file before you run the command.
What is the difference between a Claude skill and a plugin?
A skill is a folder with a SKILL.md file that Claude adds to its toolkit. A plugin is a packaged bundle that is installed through a plugin marketplace. Skills are also part of an open standard; plugins are a Claude Code packaging layer.
Can I use Claude skills in Hermes Agent?
Yes — Hermes Agent skills are also SKILL.md-based, so a Claude skill can usually be migrated by keeping the workflow, removing Claude Code-specific assumptions, and testing it on a real task. See the Hermes Agent skills guide and the migration guide for the procedure.
Take a skill you already found
If you have picked a skill from one of the directories above, try the same SKILL.md in Hermes Agent — keep the workflow, drop the Claude Code-specific packaging, and test it on one real task.
Try Hermes Agent