Author(s): Codebook Fusion

Originally published on Towards AI.

Learn the three files that make Claude Code smarter: CLAUDE.md, settings.json and SKILL.md. Real examples, setup steps, and honest tradeoffs.

For my first few weeks with Claude Code, I did something embarrassing. I typed the same instructions at the start of every session. Use pnpm, not npm. Run the tests before you say you’re done. Leave the migrations folder alone. Then I got annoyed when Claude drifted off the rules.

The article explains that Claude Code improves dramatically when you stop re-prompting rules and instead store them in three purpose-built files: CLAUDE.md (project context and guidance), settings.json (permissions and enforced hooks so Claude can’t “talk its way around” constraints), and SKILL.md (repeatable workflows you can run by name, optionally bundling supporting files). It walks through what to put in each file, key tradeoffs (e.g., guidance vs enforced rules, keeping files small, starting permissions narrow), a quick 15-minute setup checklist, common failure modes like bloating CLAUDE.md, and an FAQ clarifying where these files live and how skills work. The takeaway is to write all three files small and focused so your workflow stays consistent with far less manual repetition.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI