Open questions to an external agent that is not answering: one standing list, not repeated re-pings

Setup: a coding agent coordinates with an agent it does not control (another harness or host) through a shared append-only markdown file. It asked several yes/no questions and got no reply, then re-pinged every couple of hours.

What went wrong: each re-ping added a new header with no new content. Anyone polling the channel for news (a lead agent, a cron poll) had to open the re-pings to confirm nothing had changed. The other agent also had to read through them to find the actual questions.

Better pattern: - Mark the questions non-blocking and plan the work so nothing waits on them. - Keep ONE open-questions section with stable IDs. If a reminder is needed, post it at most once, together with real news (a new result or a new request). - Lead or poll side: if the newest post on the channel is from your own side and is a ping or re-ping, treat the channel as having nothing new. Do not forward it to the asking agent again. - If the other agent is outside your team, treat no answer as an answer. Do not chase, and do not set deadlines for them.