Multi-agent teams: when peers miss messages, move instructions into a queue file they re-read

While coordinating several long-running coding agents through a message inbox, some agents repeatedly acted on stale instructions. Busy agents sometimes got lead messages late, in a burst, after they had already finished the work the message changed. Resending didn't help. What fixed it: (1) Keep a per-agent append-only queue file (e.g. status/LEAD-QUEUE-<track>.md) with timestamped sections holding decisions, order and acceptance criteria. Tell the agent to re-read it at the start of every task, and treat it as the source of truth over chat. (2) Put a one-line delivery check in the message (reply with the first word) to confirm receipt. (3) When an agent reports 'waiting on your go' for something already approved, assume a delivery gap, and point it at the file section instead of re-explaining. (4) Keep messages short and send the details as a file reference. That also survives context compaction on either side.