Every Sunday I share one setup habit that keeps an AI agent fleet healthy for the week ahead. This week’s is about the files your agents leave behind, because I just measured mine and it was worse than I expected. This is the AI agent cleanup check I’m adding to my Sunday list.
Here’s the receipt. My agents keep two notes files: one for long-term memory and one for a running log of what each run did. Together they’re about 8.4 MB. Sitting next to them in the same folder are 470 backup copies of those two files, taking up 1.47 GB. That’s roughly 70% of the entire project folder, and it’s made of copies of 8 MB.
The part that stopped me: no instruction file tells any agent to make them. I searched every skill file and script for “bak”. Zero hits. Somewhere along the way one run decided that saving a copy before editing was careful, the habit got written into the notes, and other agents picked it up from there. Last week the fleet ran 94 jobs and made 102 backups. That’s more than one per run, and nothing reads any of them.
Why this matters even when the disk is fine
My disk is 27% full, so this isn’t an emergency. The cost shows up in other places:
⚡ GET THE AI EDGE
Weekly AI tips that actually save you time and money. No fluff, no hype — just what works.
- Search gets noisy. Any time an agent (or I) searches the folder, a single hit can show up 250 more times unless the search excludes the copies first.
- Nobody knows which copy is the right one. A backup only helps if you can pick the right version and restore it. With 251 copies of one file from 20 different agents and no restore ever tested, these are a pile, not a backup.
- It compounds quietly. The live notes files grow every day, so each new copy is bigger than the last.
The 10-minute AI agent cleanup sweep
- List what your agents leave behind. Search the working folder for leftover patterns:
.bak,copy,old,-v2,tmp. Add up the count and the size, then compare it with the files that actually matter. Big gaps are what you’re looking for. - For each pattern, ask “what reads this?” Search your agent instructions for that file pattern. If no instruction writes it and none reads it, no one designed it. The agent made it up, and nobody is responsible for it.
- Give every leftover an owner and an expiry. One line in the instructions does it: “Only the nightly job makes backups. Keep the newest 7. Delete the rest.” One backup on purpose beats 470 by accident.
- Test one restore. Pick a backup and make sure you could actually recover from it. Until you’ve done that once, it’s still just a pile.
I ran the same check on images while I was at it, expecting that to be the messy part. It wasn’t. In the last 30 days my agents generated 473 images, and only a handful of blog images were rejected and left unused. Each one was flagged by name in the run log. The mess wasn’t where agents make mistakes. It was where they do something “helpful” and never mention it.
I haven’t deleted anything yet. Deleting is hard to undo, so that call is mine, not the agent’s. If you want the backup side done properly, here’s how to make sure you can undo anything your AI agent does. This sweep also pairs well with the 10-minute memory check and with bounding what your agent reads at startup.
Want a second pair of eyes on what your agents are quietly piling up? Book a strategy session and we’ll find out together.
See you next Sunday.
Jon
📥 FREE: THE AI PLAYBOOK
The exact tools and workflows I use to run a one-person agency. 25 years of marketing experience distilled into an actionable guide. Yours free.
AI & Digital Marketing Strategist
With over 25 years in digital marketing, e-commerce, sales, and international business, Jon Jones is a seasoned entrepreneur dedicated to empowering small businesses through innovative AI solutions. His journey began in investment banking on Wall Street, guiding companies through capital acquisition and complex markets, and has expanded to include expertise in web development, brand design, product development, and courses and training.
Fluent in Mandarin and with a rich foundation in both finance and tech, Jon launched JonJones.ai to provide accessible, transformative tools for small businesses to streamline operations, generate leads, and build sustainable growth. His mastery of AI, automation, and digital strategy has enabled numerous businesses to bridge the gap between traditional practices and modern technology, ensuring they thrive in today’s digital landscape.
