CV review skill

A Claude skill that reads a CV like Karmen and includes insights from other tech recruiters.

Built at //kood for students and alumni applying to tech roles. It works on any CV, and it knows how to frame a //kood background for recruiters who have never heard of the school.

What it does

• At first it runs the 30-second test on the whole page, not line by line.
• Then it walks the CV section by section against a rubric: header, title, summary, experience, skills, education, projects, length, keyword match.
• Names the fixes in priority order, quotes your own lines back, and shows the pattern to rewrite them with.
• Matches the CV against a specific job posting, both the technical requirements and the behavioural ones.
• You can use the skill to improve cover letter or CV. Make sure you share enough context to give accurate suggestions.

Install

Pick the tool you use.

Claude.ai or the Claude desktop app

Works on every plan, including Free.

1. Download cv-review.zip. Keep it zipped.
2. In Claude, open Settings > Capabilities and turn on Code execution and file creation. Skills need it.
3. Open Customize > Skills, click +, then Create skill, then Upload a skill, and pick cv-review.zip.

That's it. Start a new chat and share your CV.

Use the zip from the link above, not GitHub's green Code > Download ZIP button. That one has a folder called cv-review-skill-main inside, and Claude needs the folder name to match the skill's name, cv-review.

Claude Code

Type these two commands into Claude Code, one at a time:

/plugin marketplace add karmenti/cv-review-skill

/plugin install cv-review@cv-review-skill

The first tells Claude Code where to find the skill. The second installs it. If the skill doesn't show up right away, restart Claude Code.

To check it worked, type /cv-review and look for it in the suggestions. You can start a review that way, or just share your CV and ask for feedback.

git clone https://github.com/karmenti/cv-review-skill.git "$HOME/.claude/skills/cv-review"

Codex, Cursor, Gemini CLI, GitHub Copilot and others

Open a terminal and run:

npx skills add karmenti/cv-review-skill -g

It detects which coding agents you have and installs the skill for them. If it asks you anything, the suggested answers are fine. -g installs it for your user account, so it works in every project. Restart your agent afterwards so it picks up the new skill.

This needs Node.js 22.20 or newer. Run node --version to check.

git clone https://github.com/karmenti/cv-review-skill.git "$HOME/.agents/skills/cv-review"

gemini skills install https://github.com/karmenti/cv-review-skill.git

To start a review, share your CV and ask for feedback. To call the skill by name:

Any other AI chat

Open SKILL.md, copy its contents into the chat, attach your CV, and ask for a review. For better results, paste references/rubric.md and references/star-method.md too.

Updating

If you installed it as a Claude Code plugin, run /plugin marketplace update cv-review-skill and then /plugin update cv-review@cv-review-skill. If you installed with npx skills, run npx skills update cv-review -g. If you installed with git clone, run git pull inside the skill folder. If you uploaded a zip, repeat the steps above with a fresh download.

How to use it

Give it two things:

1. Your CV. A PDF, a Word file, or pasted text.
2. The job posting. Paste the full text of the posting, not only the link. Some tools cannot open links, and a LinkedIn page often needs a login, so the AI may never see what the job asks for. If you have not picked a role yet, name the direction instead: junior backend, QA, frontend.

A first message can be as short as:

Here is my CV and the posting I'm applying to. What should I fix before I apply?

You get back a short verdict, a numbered list of fixes ranked by impact, and an action plan.

The feedback comes in the language your CV is written in. The skill is tested in English and Estonian.

Your data. The skill stores nothing and sends nothing anywhere. Your CV goes to whichever AI tool you run it in, under that tool's privacy terms, same as any other chat.

Which model to use

Pick the strongest model your plan offers. If the tool lets you set the effort level, medium is enough; you don't need the maximum. In our testing, Claude Opus gives consistently good reviews.

A faster, lighter model still works. Its review tends to be longer and flatter, with every point at the same weight, so you have to decide yourself what matters most.

What is in the repo

A note on examples

Every example in this repo is invented. No real CV, name, or contact detail is used anywhere in it, including the git history. The one real contact detail is the author's address in the closing line.

License

MIT, see LICENSE.

//kood, kood/Jõhvi, kood/Võru and kood/Järva are the school's names. Naming the school on your own CV is, of course, fine. The MIT license covers the text in this repository; it does not make a fork or a derived tool a //kood product.