
Give your Bots house rules so they don't freeze their computer
Cursor staff wrote a rules file that keeps several Bots from running heavy jobs at once on the one computer they share. Here's how to install it.
If you run several Bots, give them one shared rulebook for heavy work. Cursor staff member deanrie posted a ready-made file on Saturday that tells Bots to check memory first, run one heavy job at a time through a shared lock, never start virtual machines without asking, and clean up after themselves. You save it once and tell each Bot to follow it.
- Status: staff advice, posted in a support thread. It's a file of instructions, not a product setting.
- Affects: anyone whose Bots run builds, tests, installs, data scripts or browsers on their Grok Bot computer. The rules file itself says "All Bots share ONE computer with limited memory and CPU."
- Trigger: one business user's computer froze twice in two days (details below). It's one account's story, but the advice applies to anyone with busy Bots.
What happened#
A user whose Bots run parts of their business reported their computer stuck for over a day. On Saturday at 8:11 AM ET, deanrie explained that staff replaced the frozen computer on Oct 15 around 5:10 AM ET, and the new one froze "in the same way" on Oct 16 around 2:30 AM ET. "Before it froze, memory was almost completely full. There were nearly 1000 processes running, and the virtual machine running on the computer alone was using about a third of the memory." deanrie added: "This shouldn't happen. The computer should recover on its own, not stay frozen."
The user then asked for "a markdown file from you that can instruct the bots how to avoid this." At 12:33 PM ET deanrie posted one.
Why it matters#
Each Bot decides on its own when to start a build or open a browser. None of them knows what the others are running. Our read: the file gives every Bot the same habit of checking first and taking turns, which is the part a single Bot can't do alone. deanrie's summary of the pattern: "In both freezes, memory was almost full with hundreds of processes."
How to set it up#
- Copy the file from deanrie's post. Its main rules:
- Before heavy work, run
free -mandps aux --sort=-%mem | head -15. If free memory is "under ~25%" or another heavy job is running, "do NOT start." - Run heavy jobs through one lock:
flock -w 1800 /workspace/.heavy-job.lock <your command>. If the lock isn't free in 30 minutes, stop and say so. - Cap parallel work, for example
make -j2,npx jest --maxWorkers=2orpytest -n 2. - Never start virtual machines or emulators without approval, and never kill another Bot's processes without asking.
- After each task, stop leftover dev servers and watchers and close unused browser tabs.
- Before heavy work, run
- Have one Bot save it as
/workspace/RESOURCE_RULES.md. deanrie says/workspace"is the part of the computer that survives updates and recovery." - In each Bot's chat, send deanrie's line: "From now on, before any heavy task, read and follow /workspace/RESOURCE_RULES.md. Remember this as a lasting rule."
Prompt (ours), to paste with the file into one Bot's chat:
Save the file I'm pasting below as /workspace/RESOURCE_RULES.md on your computer. From now on, before any heavy task, read and follow /workspace/RESOURCE_RULES.md. Remember this as a lasting rule. Then run free -m and tell me how much memory is free right now.
Good to know#
- If the computer freezes, don't click anything. deanrie: "please still don't click Update, Recover, or Reset. While the computer is frozen, Update won't be able to finish, and Reset will roll everything back to the last save."
- Your Bots are kept apart from the computer. "Bots and their saved data are stored separately from the computer," deanrie wrote. In this case only "the last couple of hours before the freeze" were at risk.
- The chat lines matter. The file asks Bots to post a line before and after each heavy job and log it in
/workspace/resource-log.md. deanrie: "The chat lines matter more than the file." - This is a workaround. Staff say a frozen computer "should recover on its own" and have passed this case to the team. The changelog has no entry about it as of 6:48 AM ET.
What to watch#
Whether Cursor makes frozen computers recover on their own, as deanrie says they should, and whether the user's computer comes back. deanrie said "Nothing has changed on the computer yet" at 12:33 PM ET Saturday, with no newer post as of 6:55 AM ET Sunday. Follow the thread.
Steps
- Ask any one Bot to save deanrie's rules file on its computer as /workspace/RESOURCE_RULES.md (paste the file from the forum post into the chat).
- In each Bot's chat, send: From now on, before any heavy task, read and follow /workspace/RESOURCE_RULES.md. Remember this as a lasting rule.
- Our suggestion: ask one Bot to run free -m and ps aux --sort=-%mem | head -15 and show you the output, so you know what normal looks like.
- Watch for the one-line notes the rules ask Bots to post in chat before and after each heavy job.
- If the computer ever freezes, don't press Update, Recover or Reset, as staff advised.


