I sold almost everything I owned last month.
My car, a box of cables that every man inexplicably keeps in a drawer somewhere, boxing gloves I wore once, a microphone I bought when I thought I'd start a podcast (I didn't), and a pile of toys my son doesn't even look at anymore.
We moved from Cyprus to Berlin (we’re here for 5 weeks, kind of a test run to see if we like the city as a family), and the rule was: if you don't use it every week, it goes.
The process was tough. I almost felt like I was losing a part of myself.
But once most of it was gone, something unexpected happened.
Everything got clearer.
Like my brain had been quietly tracking every item I owned, running a background process I didn't even notice, and suddenly it could stop.
I'm sitting in our new Berlin apartment now, my body trying to readjust to 11-degree weather after 4 years in Cyprus (my immune system is filing a formal complaint).
I open my laptop, flick open Claude, and just stare.
It looks just like my old Cyprus place.
Digital cupboards stuffed with things I'd been hoarding for months.
Skills, prompts, instructions, context files. Most of it I'd completely forgotten was even there.
How your Claude setup gets worse over time
When you use Claude for a while, you eventually build up quite the collection of custom instructions.
Maybe Claude gave you a 400-word answer when you needed two sentences, so you added an instruction to your setup: "Be more concise."
Next week, a draft came back sounding like a corporate memo, so you added another: "Use a casual tone."
A month later, you watched a YouTube tutorial and dropped this guru’s new skill into your setup.
You keep hoarding, and it feels productive because you're fixing problems as they come up.
But 3 months in, you've got hundreds of rules piled on top of each other.
Some of them contradict each other.
"Be concise" and "always explain your reasoning" are actively fighting in the background of every response Claude gives you.
And Claude is trying to follow all of them at once, which means it's doing none of them well.
It's like handing a chef a 47-step recipe when they only need 12.
The extra 35 steps don't help. They slow the chef down, make them second-guess the parts they already know, and the dish comes out worse than if you'd just let them cook.
That's what over-prompting does to Claude.
Even the people who built Claude were giving it too many instructions
Anthropic just published a piece on how they build with Claude Code.
Their own engineering team, the people who built Claude, found that their instructions were making the output worse.
Too many instructions and too many guardrails stacked on top of each other.
If even Anthropic is over-prompting their own model, your custom setup is almost certainly doing the same thing.
The one prompt that cleans up your entire setup
So here's the move.
Instead of manually reading through your files line by line, trying to figure out which instructions are helping and which are dead weight, you tell Claude to audit itself.
If you're using Claude's desktop app, Claude already has access to your claude.md (the file where your preferences and rules live), your skills folder, your context files. Everything it needs.
Open Claude Code or Cowork and paste this:
Read my entire setup before responding. Check my claude.md, every skill in my skills folder, every file in my context folder, and any other instruction files you can find.
Then go through every rule, instruction, and preference you found. For each one, tell me:
1. Is this something you already do by default without being told?
2. Does this contradict or conflict with another rule somewhere else in my setup?
3. Does this repeat something that's already covered by a different rule or file?
4. Does this read like it was added to fix one specific bad output rather than improve outputs overall?
5. Is this so vague that you'd interpret it differently every time? (example: "be more natural" or "use a good tone")
Then give me a list of everything you'd cut with a one-line reason for each, a list of any conflicts you found between files, and a cleaned up version of my claude.md with the dead weight removed.
One message. Claude reads your entire setup, audits every rule, and tells you exactly what to cut and why.
You don't dig through files or reread every instruction yourself. It does the whole thing in about 2 minutes.
I ran it on my own setup and deleted roughly half my instructions.
And every output since has been noticeably better, because Claude has room to actually focus on the instructions that matter instead of juggling 30 rules and doing all of them at 60%.
That means less time editing outputs. Less "why did Claude ignore what I told it.”
And more of the work getting done the way you actually wanted it.
What to do after the audit
Once you get the results, don't blindly delete everything it flags.
Here's the process:
Read what it flagged and why
Delete the flagged rules
Run your 3 most common tasks with the trimmed setup
Did the output stay the same or get better? Those deleted rules were dead weight
Did something specific break? Add back just that one rule
The goal is to find your minimum viable setup.
The smallest set of instructions that gets you the output you want, with nothing extra dragging down the quality.
When I packed up our apartment in Cyprus, there was this brief window where we had almost nothing.
And in that window, I noticed which things I still actually reached for.
The Vitamix, because I use it every day. My key lights, because I record videos.
A bread knife, because I eat sourdough bread every day (this is Berlin after all, bread is basically a religion here).
Everything else was stuff I'd accumulated because it seemed useful at the time. But it was quietly taking up space, both in the apartment and in my head.
Sometimes you need to burn it all down and start from ground zero.
Your Claude setup works the same way.
Strip it down, see what you actually miss, and rebuild from there.
The rules that survive the cut are the ones that matter. Everything else was slowing you down without you noticing.
Addition by subtraction.
Btw, I'm going on the This Week in Startups podcast today (at time of writing this) and I'm kinda shitting my pants about it.
Wish me luck. (Update: it went great! Jcal said he wants to hire me haha)
See you next week,
Ole
P.S. This Skill is just the tip of the iceberg.
You can have so much automation running in your business that you get the output of a $500k/year marketing team.
That's why I'm hosting an in-depth Claude workshop.
180 people joined the last one, average review was 4.8/5.
In two hours I'll show you how to build Claude Cowork systems that handle your content, research, and workflows, so you focus on the stuff that matters.
Pre-register to reserve a spot (those on the waitlist hear first, no payment needed): https://tally.so/r/LZbxKl
