The 7 Rules That Keep Your AI Agent Running (PDF Guide)
$9.00
Why scheduled AI agents quietly stop, and how to keep yours running. A 6-page field guide: the approval switch that stalls most agents overnight, the “it didn’t run” checklist, the seven rules I put on every agent I run, and a run-rules block to paste on top of any scheduled prompt. Written for the Claude desktop app; the rules work anywhere.
Description
Why scheduled AI agents quietly stop, and how to keep yours running.
You set an AI agent to run every evening. It works for two nights. On the third morning there is nothing there, no error, and no clue why.
I run more than two dozen scheduled agents, and every rule in this guide was paid for with a run that quietly failed. You get to skip that part. Six pages, about fifteen minutes to read, and you can use it tonight.
What is inside
- The one fact that explains most failures: where your agent actually runs, and what has to be on for it to run at all.
- Setting up a scheduled task by hand in the Claude desktop app, field by field.
- The approval switch. One setting makes an agent stop and wait all night for a click nobody is there to give. I lost about 172 hours of run time to it before I understood what I was looking at. The fix takes five minutes.
- The “it didn’t run” checklist, five causes in order of likelihood, so you check the likely one first.
- The seven rules I put on every agent I run, each with the failure that taught it.
- The run-rules block: the seven rules written as instructions, ready to paste on top of any scheduled prompt.
- What to automate first, and next, in the order that keeps a mistake cheap.
Before and after
Before: a missing page in the morning, and an hour of guessing whether the computer slept, the app closed, or the agent is still sitting there waiting.
After: a log that tells you what happened, a checklist that finds the cause in a couple of minutes, and agents built so a stalled run still leaves you something usable.
“I’m not a tech person.”
You do not need to be. There is no code. It is settings to click, one block of plain-English instructions to paste, and the reasons behind each one.
What you need
The menu steps are written for scheduled tasks in the Claude desktop app on a paid Claude plan. The seven rules and the run-rules block work in any tool that runs an AI agent while you are away.
Already own The Tomorrow Briefing?
Then you have most of this. An earlier edition of this guide ships inside The Tomorrow Briefing as The Unattended-Run Kit. The Briefing is the ready-built agent with these rules already in it; this is the guide on its own, for building your own.
Who made this
I’m Robin. CMA, MBA, and a CFO for about thirty years. I write The Side Table, a monthly CFO’s read on running a business with AI. I build the things I wish somebody had handed me. This is one of them.
$9, once. Instant download (PDF). Because you get instant access, all sales of digital products are final. Having trouble downloading? Email me at [email protected] and I will get you sorted.
Keep what earns its keep.
Only logged in customers who have purchased this product may leave a review.






Reviews
There are no reviews yet.