This is one of the largest community collections of agent skills on GitHub: about 380 skills, plus agents, personas and slash commands, grouped into plugins by domain. Engineering is the biggest share (architecture, RAG, databases, CI/CD, observability, incident response, Playwright testing), followed by marketing, product, project management, regulatory and quality work, finance, and a C-level advisory set that has the agent think like a CTO or CFO. Many skills come with small Python tools, all standard library only, so nothing has to be installed with pip.
It started as Claude Code plugins and now installs into Codex, Gemini CLI, Cursor, OpenClaw, Hermes Agent, Mistral Vibe, Aider, Windsurf and several others through conversion and install scripts. It also includes a skill-security-auditor that checks any skill folder for command injection, exfiltration and prompt injection before you install it, which is a useful habit with a library this size.
It is MIT-licensed, with about 27,800 stars and 49 contributors, though most of the work is one maintainer's. At 3,869 files and about 830,000 lines, quality varies from skill to skill. The sensible way to use it is to pick the few bundles or single skills you need rather than install everything.
- Repository: github.com/alirezarezvani/claude-skills
- Licence: MIT (MIT License)
- Language: Python. Stars: 27.8K. Forks: 3,910. Last push: Aug 30, 2026.
- Scan: safe, Oct 8, 2026, commit 19392f7
Who it is for
Claude Code, Codex and Gemini CLI users who want ready-made skills for business and engineering tasks beyond coding, and people who want a large set of examples to borrow from when writing their own.
Getting started
1. In Claude Code, add the repository as a plugin marketplace
/plugin marketplace add alirezarezvani/claude-skills2. Install one domain bundle (or a single skill by name)
/plugin install engineering-skills@claude-code-skills3. Cursor and other tools: clone, then run the installer for your tool
git clone https://github.com/alirezarezvani/claude-skills.git && cd claude-skills && ./scripts/install.sh --tool cursor --target .The /plugin commands are typed inside Claude Code, not a shell. Other bundles include marketing-skills, product-skills, pm-skills and c-level-skills. Windows users should clone with core.symlinks=true, as the README explains.
Safety scan
We cloned alirezarezvani/claude-skills at commit 19392f7 on Oct 8, 2026 and ran the checks described on the GitHub Tools page: credential patterns, decode-and-execute code, install-time scripts, committed binaries, risky CI workflows, every host the code talks to, known vulnerabilities in pinned dependencies, and project hygiene. A person read every hit. This is what we found.
- The 8 secret hits are all fake: sample private keys inside the senior-security skill's secret scanner, its tests and a redaction checklist, there so the scanner has something to find. No committed binaries and no bare-IP URLs.
- A second, skill-specific pass looked for invisible Unicode, prompt-injection phrases, exfiltration hosts, credential paths, pipe-to-shell commands, long base64 and plugin hooks. The matches were security documentation, defensive guidance or test fixtures. The 2 persistence hits are comments in skillopt-sleep's install-cron.sh telling you to paste a line into crontab yourself; nothing is scheduled automatically.
- Several plugins ship hooks that run once installed (agent-memory, handoff, skillopt-sleep, playwright-pro, security-guidance). Read, they only read and write local state files such as ~/.skillopt-sleep and make no network calls.
- The README suggests piping scripts/openclaw-install.sh into bash for OpenClaw. The script only symlinks skill folders into ~/.openclaw/workspace/skills, but downloading it and reading it first costs nothing.
- No lockfiles, so no dependency advisories. 12 workflows; the one pull_request_target workflow only checks a PR's target branch and comments or closes it, without checking out PR code. Four third-party actions are referenced by version tag rather than pinned to a commit. Licence, security policy, contributing guide and code of conduct are present.
What the scanner counted
| Check | Result |
|---|---|
| Secrets | 8 candidates found and read; see the notes above. |
| Suspicious code | 2 pattern hits found and read; every one is listed under the raw findings. |
| Install-time code | 3 installer scripts |
| Committed binaries | None. |
| CI workflows | 12 workflows. 1 uses pull_request_target, none check out the pull request head. 4 of 4 third-party actions pinned to a tag rather than a commit. |
| Network hosts | 40 distinct hosts referenced from source; most often github.com, www.googleapis.com, api.example.com, consensus.app. No URLs to bare IP addresses. |
| Known vulnerabilities | No lockfile to check: dependencies are declared as ranges, so what gets installed is whatever is current on the day. |
| Project hygiene | Has security policy, licence file, contributing guide. Missing automated dependency updates, CodeQL. |
| OpenSSF Scorecard | Not scored: the project is not in Scorecard's weekly index. |
The raw findings
Every hit the scanner wrote out, with a link to the exact line at the scanned commit. Secrets candidates are redacted.
Secret candidates (8, redacted)
| Where | Rule | Match |
|---|---|---|
| engineering-team/skills/senior-security/scripts/secret_scanner.py:165 | private-key | -----B…--- (31 chars) |
| engineering-team/skills/senior-security/scripts/secret_scanner.py:174 | private-key | -----B…--- (30 chars) |
| engineering-team/skills/senior-security/scripts/secret_scanner.py:183 | private-key | -----B…--- (35 chars) |
| engineering-team/skills/senior-security/scripts/secret_scanner.py:192 | private-key | -----B…--- (37 chars) |
| engineering/skills/skill-tester/tests/test_security_scorer.py:118 | private-key | -----B…--- (31 chars) |
| engineering/skills/skill-tester/tests/test_security_scorer.py:274 | private-key | -----B…--- (31 chars) |
| productivity/handoff/skills/handoff/references/redaction_checklist.md:19 | private-key | -----B…--- (31 chars) |
| productivity/handoff/skills/handoff/scripts/redaction_linter.py:254 | github-token | ghp_aa…aaa (40 chars) |
Pattern hits (2)
| Where | Rule | Match |
|---|---|---|
| engineering/skillopt-sleep/scripts/install-cron.sh:3 | persistence | # The user copies the line into `crontab -e` if they want it. |
| engineering/skillopt-sleep/scripts/install-cron.sh:23 | persistence | # Copy the next line into 'crontab -e': |
Installer scripts (3)
- engineering/skillopt-sleep/scripts/install-cron.sh, 30 lines
- engineering/skillopt-sleep/scripts/run-sleep.sh, 60 lines
- scripts/install.sh, 276 lines
Workflows worth a look
- .github/workflows/enforce-pr-target.yml: pull_request_target
By the numbers
| Stars | 27.8K |
|---|---|
| Forks | 3,910 |
| Contributors | 49 |
| Commits | 1,499 |
| Open issues | 13 |
| Open pull requests | 18 |
| Releases | 6 |
| Latest release | v2.12.0 |
| Licence | MIT |
| Main language | Python |
| Project age | 11 months |
| Last push | Aug 30, 2026 |
| Tracked files | 3,869 |
| Lines of code | 830.9K |
| Checkout size | 36 MB |
Lines by language: Markdown 487.9K, Python 278.7K, JSON 53.2K, YAML 2,970, JavaScript 2,658, Shell 2,024.
Questions
Is Claude Skills by Alireza Rezvani free?
Yes. The library is MIT-licensed and free to install and modify. You still need an agent to run it, such as Claude Code, Codex or Gemini CLI, under that product's pricing, and loading many skills or agents uses more tokens.
Should I install all 380 skills?
No. Each installed skill adds its description to what the agent reads, and quality varies across a library this size. Install the domain bundle you need, or single skills by name, and read their SKILL.md files and scripts first. The repository's own skill-security-auditor can check a folder before you install it.
Does it only work with Claude Code?
No. Despite the name, it installs natively into Claude Code, Codex, Gemini CLI, OpenClaw, Hermes Agent and Mistral Vibe, and its conversion scripts produce rules for Cursor, Aider, Windsurf, Kilo Code, OpenCode, Augment and Antigravity.
This post is part of GitHub Tools, where every repository is cloned and scanned before it is written up. The scan is a snapshot of one commit on one day; the repository has moved on since, so check it before you install.