프로그래머가 아닌 사람을 위한 Claude Code 팁

You're a researcher digging through industry reports. Or an analyst comparing vendor proposals. Or a consultant trying to make sense of a 50-page brief. You've heard Claude Code is powerful. You installed it. You typed a few prompts. It worked... kind of. But something feels off. You're not getting what the developers seem to be getting. Here's the thing: most people using Claude Code are barely scratching the surface. I've spent the last few months chatting with non-developers at work - people from operations, strategy, and research teams who've started using Claude Code for research, document analysis, and content work. They're smart. They're motivated. But they're missing features that would 10x their productivity. This post covers the tips that made the biggest difference. 1. Your sessions don't disappear Picture this: you've spent 30 minutes building context with Claude. You've uploaded documents, explained your requirements, refined the output. Then you accidentally close the terminal. Most people think that's it. Wrong. Run this: claude --resume You'll see a list of all your previous sessions. Pick one, and you're back exactly where you left off. If you just want to continue your most recent session without choosing: claude --continue 2. CLAUDE.md is your personal knowledge bank Every time you start a new session, do you find yourself explaining the same things? "When I say 'credible source', I mean peer-reviewed journals or established industry databases." "Format all summaries with bullet points, not paragraphs." "Always check primary sources first, then secondary commentary." Stop repeating yourself. Create a CLAUDE.md file. This is a markdown file that Claude reads at the start of every session. Put it in your home directory (~/.claude/CLAUDE.md ) for global rules, or in a specific project folder for context that only applies there. What goes in it: - Definitions you use repeatedly ("credible source means...") - Formatting preferences ("always use British English spelling") - Domain context ("I work in a fast-moving industry where context matters") - Pointers to reference materials ("for reference documents, check the ~/research/sources folder") Define it once. Claude remembers forever. You can also tell Claude to look in specific folders for reference materials - PDFs, case files, templates. Instead of uploading the same documents every session, just point Claude to where they live. 3. Build repeatable workflows with agents Got a process you run often? Maybe it's: - Summarising research papers in a specific format - Extracting key points from meeting transcripts - Comparing multiple documents against a checklist You can give Claude step-by-step instructions for these tasks. But here's what most people miss: you can save these as agents. An agent is just a reusable workflow. Instead of copy-pasting the same long prompt every time, you define it once and run it whenever you need it. Even better: you can run multiple agents in parallel. Analysing five documents? Ask Claude to kick off all five simultaneously. How do you set up agents? Let's find out. 4. Ask Claude how to set itself up This one surprised everyone I showed it to. Claude Code has an internal tool - a kind of self-documentation system - that teaches it how it works. It can look up its own features, configuration options, and best practices. So if you're unsure how to: - Create new Agents - Set up MCP integrations - Configure the CLAUDE.md file correctly - Use a specific flag or feature Just ask Claude. Ask Claude directly: "How do I set up a global CLAUDE.md file?" It'll look it up and give you the correct answer. This sounds obvious in hindsight, but most people don't think to ask the tool about itself. 5. Keyboard shortcuts that save your sanity If you're spending a lot of time in the terminal, these small things add up: - Option + Left/Right: jump between words (instead of holding arrow keys) - Cmd + Left/Right: jump to the start/end of a line - Ctrl + C (twice): exit a session cleanly 6. Maybe use a better terminal The default macOS Terminal app works. But it's bare-bones. If you're using Claude Code regularly, consider switching to something like Warp (warp.dev). It gives you: - Tabs (run multiple sessions side by side) - Clickable cursor positioning (click where you want to edit, instead of arrowing over) - Better text selection and copy/paste - Generally less friction It makes the experience noticeably smoother, especially if you're not from a developer background. Conclusion Claude Code isn't just for developers. It's for anyone who needs to gather, synthesise, and transform information - then apply their own judgement. The key word there is "judgement." Don't ask Claude to make decisions for you. It can't think through edge cases the way you can. It doesn't understand your specific context the way you do. But it can get you to the strategic work faster. It can do the legwork - reading, summarising, formatting, comparing - so y