398 Claude skills
A free library of ready-made skills, agents and slash commands for Claude Code, Codex, Gemini CLI, Cursor and nine more coding agents. What a skill is, what this repo gives you across 20 domains, and a step-by-step install for each tool.
- TL;DR and key numbers
- Skills, from zero
- The problem: a smart agent with no playbook
- What's inside the repo
- The 20 domains
- 13 tools, one library
- Setup guide, step by step
- So, is it really a khazana?
- Which bundle should you install?
- Caveats and safety
- FAQ
- Glossary and sources
TL;DR
- A skill is a folder with a
SKILL.mdfile: instructions an AI coding agent loads only when a task needs them. It turns a general model into one that follows a specific playbook (an SEO audit, a sprint plan, a DCF model). - alirezarezvani/claude-skills is an MIT-licensed GitHub repo with 398 production-ready skills, plus 123 agents, 7 personas and 157 slash commands, across 20 domains: engineering, product, marketing, compliance, finance, business operations and more.
- It works with 13 AI coding tools. Claude Code installs it natively as plugins (two commands); Codex, Gemini CLI and others have install scripts; Cursor, Windsurf, Aider and friends get converted copies.
- Don't install everything. Pick the bundle for your work, try a few skills, and read any skill before you trust its scripts.
Skills, from zero
A coding agent like Claude Code starts every session knowing a lot in general and nothing about your way of doing a task. You can paste instructions into the chat each time, or put them in a skill once.
In Anthropic's words, a skill is "a SKILL.md file containing instructions that extend what Claude can do". The file has a short header (front matter) with a description, then the instructions. Only the descriptions sit in the agent's context all the time; the full skill loads when it's used, so hundreds of skills don't fill up the context window.
Where skills live in Claude Code
| Scope | Path | Loads in |
|---|---|---|
| Personal | ~/.claude/skills/<name>/SKILL.md | All your projects on this machine |
| Project | .claude/skills/<name>/SKILL.md | That repository (commit it to share with your team) |
| Plugin | <plugin>/skills/<name>/SKILL.md | Wherever the plugin is enabled |
Claude uses a skill on its own when your request matches its description, or you call it directly by typing / and its name. Skills that come from a plugin are namespaced, /plugin-name:skill-name, so they never clash with your own.
The problem: a smart agent with no playbook
Ask an agent for "a launch plan" or "a compliance gap check" and you get a reasonable, generic answer. What you usually want is the checklist an experienced person would follow, the right frameworks, and numbers worked out by a script rather than guessed.
Without skills
- Long prompts, retyped every session
- Generic structure, different every time
- Maths done "in the head" of the model
- Writing good skills yourself takes days per domain
With a skill library
- The playbook loads only when the task needs it
- Same steps and output format every time
- Bundled Python tools do the calculations
- Hundreds of skills, already written and organised
What's inside the repo
The README describes it as "the most comprehensive open-source library of Claude Code skills and agent plugins". Each skill is a package of three parts:
SKILL.md
Structured instructions, workflows and decision frameworks.
Python tools
CLI scripts, standard library only: no pip install.
References
823 templates, checklists and domain knowledge files.
Plugin
Bundled for Claude Code's plugin marketplace.
Skills, agents and personas
| Skills | Agents | Personas | |
|---|---|---|---|
| Purpose | How to do a task | What task to do | Who is thinking |
| Scope | One domain | One domain | Cross-domain |
| Example | "Follow these steps for SEO" | "Run a security audit" | "Think like a startup CTO" |
| Count | 398 | 123 | 7 |
The table follows the README's own "Skills vs Agents vs Personas" comparison; counts are its badges.
The 20 domains
"398 production-ready skills across 20 domains", from the README's Skills Overview. To scale, by number of skills:
A few examples from the table
- Product (17): product manager, agile PO, UX researcher, landing pages, SaaS scaffolder, code-to-PRD.
- Marketing (50): eight pods: content, SEO + AEO (getting cited by AI answer engines), CRO, channels, growth, sales.
- Productivity (12):
capture(brain dump to actions),weekly-review,deep-work,meetings,roast(pressure-tests an idea). - Compliance OS (9): controls, evidence and audit-readiness workflows.
- Business Operations (7): process mapping, vendor management, capacity planning, internal comms.
- Commercial (8): pricing strategist, deal desk, partnerships, RFP responder, forecasting.
- Finance (8): financial analyst (DCF, budgeting), SaaS metrics coach, month-end close.
13 tools, one library
Skills follow the same SKILL.md format (the agentskills.io standard), so one library can serve many agents. The README lists these:
Native vs converted. Claude Code uses the repo as a plugin marketplace. Codex, Gemini CLI, OpenClaw, Hermes and Mistral Vibe read the skills through their own install scripts. Cursor, Aider, Windsurf, Kilo Code, OpenCode, Augment and Antigravity get copies converted to their own rule formats by scripts/convert.sh.
Setup guide, step by step
Pick your tool below. Each install is a few commands; none of them need an account or an API key beyond the agent you already use.
Claude Code (recommended)
- Add the marketplace. Run this inside Claude Code (it's a slash command, not a shell command):
/plugin marketplace add alirezarezvani/claude-skills
- Install the bundle for your field. Each bundle is a plugin full of skills; start with one:
# pick one (or more)
/plugin install marketing-skills@claude-code-skills
/plugin install engineering-skills@claude-code-skills
/plugin install engineering-advanced-skills@claude-code-skills
/plugin install product-skills@claude-code-skills
/plugin install pm-skills@claude-code-skills
/plugin install finance-skills@claude-code-skills
/plugin install business-growth-skills@claude-code-skills
/plugin install c-level-skills@claude-code-skills
/plugin install ra-qm-skills@claude-code-skillsSingle tools are plugins too, for example /plugin install pw@claude-code-skills (Playwright testing), self-improving-agent, code-to-prd or grill-me. The marketplace lists 104 plugins in all (its marketplace.json, read 27 Oct 2026).
- Use a skill. Just ask; Claude picks the skill whose description matches. To be explicit, name it:
Using the seo-audit skill, audit the landing page in ./site and list the top 5 fixes by impact. Using the senior-architect skill, review this service's architecture and name the top 3 scalability risks.
Type / in Claude Code to see the installed skills; plugin skills appear as /plugin-name:skill-name.
>_ OpenAI Codex
Install one category at a time: Codex fits every skill description into a fixed budget, and with a few hundred skills it drops the ones it needs to choose from (the repo's install guide). Skills go to ~/.codex/skills/.
git clone https://github.com/alirezarezvani/claude-skills.git
cd claude-skills
# no arguments lists the categories
./scripts/codex-install.sh --category engineeringGemini CLI
git clone https://github.com/alirezarezvani/claude-skills.git cd claude-skills ./scripts/gemini-install.sh
Skills install to .gemini/skills/; inside Gemini CLI, turn one on with activate_skill(name="senior-architect").
Cursor, Windsurf, Aider and the rest
Convert the skills to your tool's format, then install them into your project:
git clone https://github.com/alirezarezvani/claude-skills.git cd claude-skills # 1. convert every skill for every tool (~15 seconds) ./scripts/convert.sh --tool all # 2. install into your project (cursor, windsurf, aider, kilocode, opencode, augment…) ./scripts/install.sh --tool cursor --target /path/to/your/project
Cursor gets .mdc rules, Aider a CONVENTIONS.md, Windsurf .windsurf/skills/. Antigravity installs to ~/.gemini/antigravity/skills/ with --tool antigravity.
Just one skill, by hand
Every skill is a plain folder, so you can copy a single one into your personal skills folder:
git clone https://github.com/alirezarezvani/claude-skills.git mkdir -p ~/.claude/skills cp -r claude-skills/marketing-skill/skills/seo-audit ~/.claude/skills/
On Windows: clone with git clone -c core.symlinks=true … (with Developer Mode on) so the per-tool folders check out as real skills, and set PYTHONUTF8=1 so the Python tools don't crash on Unicode output (the repo's Windows notes).
Check it's working
- In Claude Code, type
/: the new skills should be in the list. - Ask for a task in that skill's area and watch for Claude saying it's using the skill.
- Run one of a skill's scripts with
--help: the README says every CLI entry point is verified to run that way.
398 skills for free is a lot of work you don't have to do. But the value is in the five or ten skills that match your actual job, not in having all of them loaded.
What you get
- Playbooks for work your agent does every day
- Scripts that compute instead of guess
- One library across Claude Code, Codex, Gemini, Cursor…
What it isn't
- Not an official Anthropic product
- Not a substitute for knowing your domain
- Not "install all and forget": pick and read
Which bundle should you install?
| You are… | Start with | Why |
|---|---|---|
| A developer | engineering-skills | Architecture, frontend, backend, QA, DevOps, security |
| Building with agents | engineering-advanced-skills | Agent designer, RAG architect, MCP builder, CI/CD |
| A founder or PM | product-skills, pm-skills | PRDs, discovery, roadmaps, sprints |
| Marketing your work | marketing-skills | Content, SEO, CRO, launch and growth |
| Running the numbers | finance-skills | Financial analysis, SaaS metrics |
| A student | engineering-skills + one single plugin | Learn how good skills are written, then write-a-skill to make your own |
Caveats and safety
- Skills are code and instructions you run. A skill can tell your agent what to do and ship scripts it runs on your machine. Read a skill before you rely on it. The repo includes a scanner:
python3 engineering/skills/skill-security-auditor/scripts/skill_security_auditor.py /path/to/skill/returns PASS / WARN / FAIL. - Community project. It's built by Alireza Rezvani and contributors, MIT-licensed, and not affiliated with Anthropic.
- The numbers move. The repo is updated almost daily. The README's own figures disagree in places (706 vs 766 Python tools; 345 skills in the converter section vs 398 in the header), and its "5,200+ GitHub stars" line is out of date: GitHub showed 28.7k on 27 Oct 2026. This guide uses the README header and badges.
- Context isn't free. Descriptions of every installed skill stay in context. Installing hundreds at once can crowd out the ones you need, which is exactly why the Codex guide says to install by category.
- Paths changed. Some README examples point to old folder paths; skills now sit in a
skills/subfolder (for examplemarketing-skill/skills/seo-audit). The commands in this guide use the current paths.
FAQ
Is it free?
Yes. The repo is MIT-licensed. You still pay for whatever agent you run it in (your Claude, Codex or Gemini plan).
Do I need Python?
Only for the bundled scripts. They use the standard library only, so there's nothing to pip install.
Will it slow Claude Code down?
A skill's full body loads only when it's used; only short descriptions stay in context. Installing one or two bundles is the sensible middle ground.
How do I update?
Plugins update through Claude Code's plugin system. For cloned installs, git pull and re-run the install script. The project says it keeps patch releases backwards-compatible.
Can I make my own skill?
Yes: a folder with a SKILL.md (front matter + instructions) and optional scripts/ and references/. The repo even has a write-a-skill plugin for it.
How do I remove a bundle?
Uninstall the plugin from Claude Code's /plugin menu, or delete the copied folder from ~/.claude/skills/.
Glossary
- Skill
- A folder with a
SKILL.md: instructions (and optional scripts) an agent loads when a task matches. - Front matter
- The short header at the top of
SKILL.md; its description tells the agent when to use the skill. - Plugin
- A package of skills, agents and commands that Claude Code installs from a marketplace.
- Marketplace
- A list of plugins Claude Code can install from; this repo is one.
- Agent
- Here: a pre-set specialist role that decides what task to run.
- Persona
- A cross-domain identity ("startup CTO") with its own skills and style.
- Context window
- Everything the model can see at once; skill descriptions take a little of it.
- github.com/alirezarezvani/claude-skills. README, INSTALLATION.md and
.claude-plugin/marketplace.json, read 27 October 2026. - GitHub API for the repo: 28,769 stars and 3,994 forks on 27 October 2026.
- Anthropic, "Skills" in the Claude Code docs. What a skill is, where skills live, how they're invoked.
