Development & Engineering / workflow case

Started using Codex goals and codex steering Claude code to write the code and audit the code.

Beginner to intermediate Set up once, then iterate continuously @makash
Result

Using Codex goals + Claude Code + Pi agent to collaborate on VPS migration: after a throttling failure, plan R2 backup, restic encrypted backup, and restore to a new server.

For

People who use multiple agents to run real ops migrations and want to see how the fault and recovery closed loop can be handed off to agents for assistance.

Started using Codex goals and codex steering Claude code to write the code and audit the code.

Using two virtual private servers.
Lasted 4 days before both providers throttled them.

Ranted with Claude web.

It said move to a dedicated server.

So I used pi agent to plan migration.

Since servers was badly throttled.
Made a plan to backup to cloudflare R2.
Restore from backup on the new server.

Used restic to backup. restic does local encryption. With a key.
Key very important to have at restore.

Took almost 25 hours as servers were crawling.

Got some folders restored and working.

Deleted the servers.
Because backup and restore worked.

Today decided to restore everything.

Pi says it can’t reach the older server.

I’m like:

“Why the F do you want the older server?”

Pi says:

“It is not secure to keep credentials around, so I didn’t save the keys.”

Me:

“WTF. I deleted the servers already.”

Then Pi suddenly starts restoring.

I was too scared to ask.
Finally I gathered courage.

“Where did you get the keys from?”

Pi:

“Found them in /tmp. They were still there. Do rotate them once done.”

I am now going to lie down.

YOLO.

Related