CHEAPUNO guide

Should I use a Claude Code skill file or CLAUDE.md for reviews?

Decide where to store code review preferences in Claude Code by weighing skill files against project configuration.

What you are probably seeing

You notice that review feedback changes when you start a new session because the assistant does not remember your specific style preferences.

You see that writing instructions directly in chat gets tedious as the project grows and you repeat the same correction notes.

Why it happens

Claude Code loads CLAUDE.md automatically for every session, making it ideal for constant project-wide rules that must always apply.

Skill files are only triggered when explicitly invoked, which suits optional or specialized workflows that do not need to run on every single interaction.

What changed recently

Step by step

1. List the review preferences that must never be missed, such as mandatory variable naming conventions.

2. Place these mandatory rules in the CLAUDE.md file to ensure they load automatically in every session.

3. Identify specialized checks that only apply when you explicitly request a deep code audit.

4. Create a dedicated skill folder for these optional deep-dive reviews to keep the default context lean.

5. Test both locations by starting a new session and verifying which rules apply by default and which require manual activation.

Ways to go further

Affiliate links: CHEAPUNO may earn a commission if you buy through them, at no extra cost to you.

Other problems readers raise

Does your AI coding tool handle specific language idioms well?

How to compare code review tools for maintenance and design?

How to test AI code reviews for GitHub traceability gaps?

Is technical writing as valuable as coding or AI?

Is technical writing a standalone career or just an engineering bonus?