Make Claude Code give you answers, not essays transcript

Here's how you can make your clawed code output or really any agent output a lot more readable and easier to work with. Recently, I've been talking about this extra little bit that I add to prompts saying, "Be extremely concise. Sacrifice grammar for the sake of concision. This turns large prompts into really nice human readable format." And so I thought to myself, why don't I add this into my Claude code global claude.md file?

Because then I'll get much more concise feedback coming from Claude. So I decided to do a test. I found an error that I needed Claude to debug for me and I ran it without the extra command to sacrifice its grammar. We can see that straight away it starts outputting kind of paragraphs for me. This is still pretty short and easy to scan, but it's still not as clean as it could be. I think in the concise version, I get ESM CGS mismatch.

Let me check the build output and config. Very direct. It starts using nice little headers, too. Root cause found. React 19 incompatibility with ink 4.1.0. Here's another one. Still same error, but the biggest difference is in the final summary. This is without the concise instruction where it gives me a fair chunk of stuff, a little mini article to read here. Whereas in the concise version, it really is just a few words here.

A couple of headings, a little list, and we get the information faster. Now, of course, if I need to follow this up with extra questions, I can. Under the hood, the LLM still has the same understanding of what it's done, but I prefer this concise summary as the first thing I receive. And here's what I have in my claw. MD file. When reporting information to me, be extremely concise and sacrifice grammar for the sake of concision.