Pinned Tweet
My free safety gate makes Claude ask before it posts, emails, pays or deletes.
The kit is the rest of my system. Claude starts every session knowing my business, and Monday's decisions are still there on Friday.
$39 one time. 14-day refund. I don't write code.
Link below 👇
Link-in-bio tracking: Tips for a one-person business:
Give every link in your bio its own tag, a short label in the link that names the app.
All 5 of mine now land on the page that sells, so I can see which app actually sends people.
Hope this helps! 🙂
Here's the page they all land on: sunabuilds.ai/os?utm_source=…
My starter kit for Claude Code. $39 one-time, a working result in 15 minutes.
🚨🚨HEADS UP IF YOU USE CLAUDE CODE🚨🚨
v2.1.318 fixed a bug where keys typed right after sending a message could answer a permission prompt, even a slow one.
Update today!!
It's also why my big moves, push, deploy, send, pay, delete, go through a gate that stops Claude first.
To update, run claude update in your terminal. On the desktop app, just update the app.
The gate I use is free and open source, so you can read every line first: sunabuilds.ai/gate?utm_sourc…
Every new Claude Code session used to start with me explaining my business from the top.
Now Claude starts every session knowing it, and Monday's decisions are still there on Friday.
I turned it into a kit. $39 one time. 14-day refund. Built in plain English.
Link below 👇
Here it is: sunabuilds.ai/os?utm_source=…
First result takes about 15 minutes. If it isn't for you, email me within 14 days and the $39 goes straight back to you.
Do you use Claude Code inside Cursor? Check today.
Tampered builds, (versions 2.1.312 and 2.1.313), sat on Cursor's marketplace for about 11 hours on Oct 29.
Cursor's notice has the check and the cleanup. Uninstalling alone may not be enough.
Cursor's notice, with the check for Mac and Windows: cursor.com/blog/claude-code-…
The driving factor that separates a human that is productive with AI in their life and a human that isn't productive with AI in their life, is taste and judgment.
My agent kept tripping on the same dumb thing every few days.
so now every mistake goes in a log with exactly one fix attached. second time it happens? That’s a design problem.
The agent didn’t get smarter. the system did. get a ratchet. 🙂
Here it is: sunabuilds.ai/os
First result takes about 15 minutes. If it isn't for you, email me within 14 days and the $39 goes straight back to you.
What happens when Claude goes to delete a file on my computer.
The gate stops it and asks me first. I say no, the file stays.
It does the same before pushes, deploys, posts, emails and payments. 53 tests, all passing.
Free and open. Link in the reply 👇
Here it is, code's open so you can read every line first: sunabuilds.ai/gate
suna | building in public retweeted
Replying to @nosunaaa
Helpful tip:
Before you give your AI more access, write down the actions you'd never let a new employee do alone on day one - that list is your safety gate.
Mine makes Claude stop first, and ask me before it pushes any code, deploys, posts, sends email, pays or deletes files 💯
My AI can't post, email, deploy, spend money or delete files without asking me first.
It's one script I built - 160 lines, which runs before every action. If it breaks, everything stops.
I'm giving it away free. Link in the reply 👇
Here it is, code's open so you can read every line first: sunabuilds.kit.com/6580c7b9b…



