Module send_message_to_peer 

Agent-loop tool that sends a message to a configured peer on a shared channel.

Validates the target against crate::peers::ResolvedPeers for the calling agent on the requested channel: peers must mutually opt in via a [peer_groups.<name>] block whose agents lists both, OR appear on the group’s external_peers list, before this tool will deliver. Cross-channel sends from outside the resolver’s authorization surface are rejected.

Delivery splits by target type:

• Agent-alias targets route in-process via crate::agent::loop_::process_message: alpha calls send_message_to_peer(target = "beta", ...) and beta’s agent loop runs the message. The two agents share the channel’s bot identity, so an outbound to the channel would loop the bot’s own handle back through inbound; the in-process path avoids that and lets the orchestrator deliver beta’s reply (if any) through the same channel beta is configured on.

  This path is fire-and-forget: the recipient runs on a detached tokio::spawn, so the sender’s ToolResult.success = true means “accepted for processing”, not “completed”. Recipient errors do NOT surface to the sender; they are emitted via tracing::warn! inside the spawned task and via the recipient agent’s own observability (audit log, runtime trace, channel reply). Observers diagnosing a missing peer message should look at the recipient’s spans, not the sender’s tool output.

• External peers (humans, external bots) route through crate::cron::scheduler::deliver_announcement with the external username as the platform target. The channel registry the binary registers at startup forwards the send to the live channel instance. This path is synchronous: the deliver_announcement future resolves before the tool returns, so a success = false here genuinely reflects a delivery failure.

Structs

SendMessageToPeerTool

Send a message to a peer on a shared channel. Bound to a single calling agent’s alias; the tool validates every send against that agent’s resolved peer set.