As a start to using Claude, I’ve been trying to make the most of its features. While talking with a friend, we started discussing skills.

In the Claude docs, skills are more personalized agents with a bit more context. For example, before asking the chat to create a new report, you can say: “Act as a nutritionist with a cutting goal and create the best regimen for me.” After that, the agent will be more accurate for this objective.

But if you want to reuse that context for future tasks, skills are your best option.

For different objectives, it’s possible to create a single command to reuse the context.

At work, we have some issues with RuboCop because the project has some legacy code, and it often needs to be refactored to follow RuboCop rules. This is very boring, so I created a skill to handle that for me before we send the code to GitHub.

So, I added this command to my Claude terminal session to create a skill:

Create a /review-code skill to review Ruby and JavaScript code.
Read the project's .rubocop.yml and .eslintrc.js files and extract the configured rules. Build the skill with:

Instructions to identify what to review: if no argument is passed, run git diff main...HEAD --name-only and read each relevant file before reviewing
Ruby section with rules grouped by category (Lint, Security, Metrics, Naming, Style, Layout) — separating what blocks merge from what is a warning
JavaScript section with rules grouped (Errors, Warnings, Modern code in app/webpack/, Vue, Legacy code in app/assets/javascripts/)
Report format with: 🔴 Blockers, 🟡 Warnings, 🟢 Good practices observed, 📋 Summary

Be direct and specific: quote the problematic snippet and show how it would look corrected. Do not list rules that were not violated.

After that, the command will be available in your Claude project session. You can invoke it by typing /, as shown below: