My Brain Dies at 2PM. I Built an Obsidian + Claude Second Brain
I built a Claude Skill that reads my Obsidian vault and answers client messages when my brain is foggy. Full setup, CLAUDE.md, and 5 brain-off prompts.
According to Upwork, I worked 2,106 hours in the last 12 months.
That’s 8.3 hours a day. Every day.
On top of that, I run LearnAIWithMe with my wife.
Articles, vault, newsletters, experiments with AI tools, and another 3-4 hours per day. (at least.)
So by 2 PM some days, my brain just stops.
I stare at a client message for ten minutes, and nothing happens.
The words are there. The meaning is not.
If you have ADHD, you know this.
If you don’t, but you work a lot, you know this too.
I got tired of losing afternoons.
So I built something.
TL;DR: I built a Claude Skill called problem-solver. It reads my Obsidian vault, takes any panicked client message, and hands me back a calm to-do list in 20 seconds.
Full code, CLAUDE.md config, vault structure, and 5 brain-off prompts below. Setup takes under an hour.
The system in one sentence
Every client conversation, meeting, decision, and Slack thread lives in an Obsidian vault.
Claude Code reads that vault.
When my brain stops, Claude picks up.
Three months ago, I wrote about the basic setup.
This post is what I built on top of it: a Skill that takes any client message and hands me back a calm to-do list, even when I can’t think.
What it looks like when it’s working
Let me show you with a demo.
“Acme Corp” isn’t a real client, real client data stays private, but the vault, the Skill, and the workflow are exactly what I use every day.
I mocked up Acme’s folder with realistic meeting notes, Slack logs, and a project brief. Then I ran a real client-style message through it.
Here’s the message.
“Hey, we need to revisit the pricing page before Friday. Also the Stripe webhook is still firing twice on renewal. And are we still on for the Tuesday call or did that move? Let me know.”
Three questions in one message.
Your brain reads “Friday” and panics if you’re burned out or already worked 7 hours that day.
Old you would stare at this for 20 minutes.
New you paste it into Claude Code and say: use problem-solver skill, client is Acme Corp.
Claude goes to Acme Corp/ in the vault. Reads the last three meeting notes, the current month’s Slack log, and the project brief. Comes back in 20 seconds.
Here’s what I get back.
A breakdown of what we already know, what was decided, and what to do next.
The Skill does the same on your real client conversations, once your vault has them.
When you don’t have the energy, the to-do list acts in your place.
Why does this work when I can’t think
The skill does three things my tired brain can’t:
It separates signal from urgency. The message felt like a fire. It wasn’t. It was three normal items stacked with a Friday reference that my amygdala heard as “emergency”.
It remembers what I decided last week. Half the panic in client messages is “did I already answer this and forget?” The vault knows. I don’t have to.
It caps the to-do list at six items, each under 15 minutes. An overwhelmed brain can’t do a 3-hour task. It can do a 12-minute task. The Skill enforces this rule, so I don’t have to.
The vault is the memory.
The Skill is the translator between “panicked client message” and “thing I can actually do right now”.
Below the paywall: the full Skill code you can drop into Claude Code today, the vault structure that makes it work, the exact CLAUDE.md config, and the five brain-off prompts I use when I can barely form sentences.
The Setup
Four pieces. Each one does one job.




