The cockpit showed six agents, all orphaned. ps showed six agents, all alive, happily working. Nothing had crashed. The bug wasn’t in my code — it was in which tmux server my code was talking to. Three isolation footguns, and how each one bites.
I restarted the daemon and the cockpit came up with an empty fleet. tmux still had every agent, alive and mid-task. The append-only store had lost nothing — but it loaded zero records, because a segment rotation had corrupted the index over data that was sitting right there intact.
I kept spawning more coding agents than I could track. So I built warden — a single Go binary that spawns, watches, and prices a whole fleet from one cockpit, whatever agent backend you run.
