Cross-session agent messaging connects agents that operate in separate conversation or execution sessions. They can announce changed files, request help, share findings, report blockers, or coordinate the order of dependent work.
A reliable channel identifies the sender, target, task, and current state without treating untrusted content as commands. Messaging should complement clear file ownership and synchronization rather than attempt to conceal conflicting concurrent edits.
ELI5
Cross-session agent messaging lets AI agents in separate work sessions send each other updates or requests. The messages can coordinate related work even though each agent has its own conversation or execution environment.
For example, one coding agent can tell another that it changed a shared interface before the second agent edits a dependent file. Each message should identify the sender, task and current state, and untrusted message content must not automatically become an instruction.
