20 Lakh Free AI Agent Skills — The 5 to Install First
20 Lakh Free AI Agent Skills — The 5 to Install First
AI agents get powerful when you give them skills. A skill is just a SKILL.md file — instructions that teach the agent how to do one job really well.
Here is the crazy part: there are now over 20 lakh free skills on the internet. All open source. All copy-paste.
This guide shows you where to get them, and the 5 to install first.
Where the Free Skills Are
Two places cover almost everything:
1. skillsmp.com — a search engine for skills. It indexes 2,124,366+ open-source SKILL.md files. That is more than 21 lakh. All free. Search anything — “kubernetes”, “resume”, “sql” — and copy the skill.
2. github.com/alirezarezvani/claude-skills — a curated repo with 337 ready skills. MIT licensed. Over 20k stars. This is the best starting point because every skill is tested and organised.
The 5 Skills to Install First
Do not install 50 skills. Start with these 5. They cover the work every engineer does daily:
- pr-review-expert — reviews your pull requests like a senior engineer. Finds security bugs, bad patterns, missing tests.
- api-design-reviewer — checks your API design before you build it. Naming, status codes, versioning, pagination.
- code-review — a general code reviewer for any language. Great for college projects before submission.
- docs-writer — writes clean READMEs and docs from your code. Saves hours.
- test-generator — generates unit tests for your functions. Perfect when you “forgot” to write tests.
How to Install (One Command)
Installing a skill = copying its folder into .claude/skills/ in your project. That is it.
git clone https://github.com/alirezarezvani/claude-skills
cp -r claude-skills/engineering/skills/pr-review-expert .claude/skills/Restart Claude Code. The skill is live. Repeat for the other four.
Skills make agents smart. Labs make YOU smart.
Installing skills takes 2 minutes. Knowing what good code looks like takes practice. DeployU gives you hands-on labs where you build and deploy real projects — the stuff reviewers actually check for.
We Tested It: pr-review-expert vs a Sample API
We took a small sample Express API — the kind students build for college projects — and ran pr-review-expert on it.
In one pass, it found:
- SQL injection — a query built with string concatenation. It flagged the exact line and rewrote it with parameterised queries.
- Hardcoded credentials — a database password sitting in the source code. It suggested moving it to environment variables, with the exact code change.
- Missing auth — a delete endpoint anyone could call. It added a proper auth middleware check.
Every finding came with the fix, not just a warning. A human senior review like this takes a day. The skill did it in about a minute.
Your Turn
Clone the repo. Install the 5 skills. Run pr-review-expert on your own project.
You will probably find the same three bugs. Better you find them than your interviewer.
Want projects worth reviewing?
DeployU gives you production-grade projects on AWS, Azure, and GCP — real deployments you can put on your resume. Built for Indian engineering students.
