
Claude Skills Collection
380 MIT-licensed agent skills across engineering, marketing, product, and ops — cross-agent compatible.
About
alirezarezvani/claude-skills is one of the largest community agent-skill libraries: around 380 skill packages plus 30+ agents and 70+ custom commands, spanning engineering, marketing, product, compliance, C-level advisory, research, business operations, and finance. It uses the same open SKILL.md specification Anthropic published, and its cross-agent design lets the same packages run natively on 13 AI coding agents with conversion scripts handling per-agent format differences.
The repo has attracted roughly 28,000 GitHub stars since its October 2025 creation, making it one of the most-starred skills collections. It ships with a built-in security auditor skill and CI quality gates, and its Python tooling uses only the standard library. Progressive disclosure keeps context usage low: agents read only each skill's name and short description until a request triggers the full skill.
Limitations: it is a community project, not affiliated with Anthropic — the name can mislead. With 380 skills, quality varies and curation matters; install selectively rather than everything. As with any third-party skill library, review SKILL.md files and scripts before installing since skills run with the agent's full permissions.
What it can do
- 380 skill packages across engineering, marketing, product, ops
- Cross-agent compatibility (13 coding agents)
- Built-in security auditor skill and CI quality gates
Setup
Clone https://github.com/alirezarezvani/claude-skills and install selected skills per the README, or use npx skills add with the repo.
Prerequisites
A skills-compatible coding agent; review skill files before installing.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 60; Trending score 88/100 from researcher assessment.
★ 28,071 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.