Strip away the Unreal skills and the MCP, and Claude is still probably the best general-purpose colleague you've ever had: a writer, editor, analyst, researcher, and tutor that's available at 2am and never annoyed by a dumb question. Most of your week isn't in the engine. This page is about that part.
The habit to build: before you open a blank doc, a blank email, or a gnarly spreadsheet, spend thirty seconds telling Claude what you're trying to do. Worst case, you get a mediocre first draft to react to — and reacting is far faster than staring at blank. Best case, it just… does the thing.
If explaining the task takes 30 seconds and doing it takes 30 minutes, delegate it. That trade is available to you dozens of times a week now.
Claude never judges, never gossips, never gets tired of explaining. The question you've been embarrassed to ask for two years? Ask it today.
Claude's output is a strong second-to-last draft. Your judgment, taste, and name go on the final. That division of labor is the whole trick.
First drafts from bullet points, tone rewrites ("make this firmer without being rude"), tightening a rambling doc to half its length, or pressure-testing an important message before you send it. Give it your raw thoughts, not a blank page — Claude is best as an amplifier of what you actually mean.
Before: "Here's the agenda and last week's notes — what should I make sure gets decided?" After: transcript in, decisions and owned action items out. You'll never again leave a meeting wondering what actually got agreed.
Drop in a CSV or spreadsheet and ask questions in English: find the outliers, chart the trend, explain why the totals don't match, write the formula you'd otherwise Google for twenty minutes. Claude does the actual analysis — not just advice about how you could do it.
"Summarize this 40-page GDC talk transcript," "compare these three middleware options for our use case," "what changed in the new platform TRC?" Claude reads the long thing so you read the one-pager — and it cites where claims came from so you can spot-check.
The fastest way to grow into an adjacent discipline is sitting in your terminal. Artists learning what makes shaders expensive, designers learning to read Blueprint, engineers learning lighting vocabulary — ask for an explanation at your level, then ask follow-ups until it clicks.
Twenty name options for the new ability. The three strongest arguments against your own plan. A pro/con table for a decision you've been circling for a week. Claude is a tireless sparring partner — and explicitly asking it to argue against you is one of the highest-value prompts there is.
Who it's for, what you're trying to achieve, what you've already tried, what good looks like. Vague in, vague out — same as delegating to a human.
"Write it like this one" with a sample attached outperforms three paragraphs of description, every time.
The second message is where the magic is. "Good, but shorter, and cut the third point" — treat it like a conversation with an editor, not a slot machine.
"Explain why you chose this approach" catches misunderstandings early — and teaches you something in the process.
Claude takes correction gracefully and adjusts immediately. And if the correction is something the whole team should know — into the skills it goes. See exactly how →
In the repo? Use Claude Code — it can see the project, the skills, and the MCP. Writing, research, spreadsheets, thinking? The Claude app is faster to reach for and just as capable. When in doubt, either works — they're the same teammate wearing different jackets.