[Bug]: Bot-to-bot (message_bot) replies fill the user's 1:1 chats; fold them under the 'Messaged X' marker #1366

Confirmation

• This is not a security vulnerability report

Description

When one bot messages another with message_bot, the woken bot runs with trigger bot_message inside its 1:1 thread with the user. Its final reply (or the automatic "Update from : …" fallback when it stays silent) is stored as a normal message in that thread. The sending bot's thread also gets follow-up text about the exchange.

The thread already shows a clickable "Message from X" / "Messaged X" marker (CollaborationMarker) for the exchange, so the extra text duplicates it. With a few bots coordinating, the user's DMs fill up with bot-to-bot chatter and it's hard to see what actually needs them.

Suggestion: in a bot's 1:1 thread, fold output from bot_message runs under the existing marker (collapsed by default, expandable), while keeping anything that needs the user visible (asks, choices, approvals, takeover). Display-only: stored history, model context and exports stay unchanged. Ideally the same on web/desktop and the native mobile app (apps/mobile/lib/message-presentation.ts).

Steps to reproduce

1. Create two bots, A and B.
2. In A's 1:1 chat, ask A to check something with B (A calls message_bot).
3. B replies to A.

Expected behavior

A's and B's chats with the user show the "Messaged B" / "Message from A" markers; the bot-to-bot content is available by clicking the marker, not as full messages in the user's DM.

Actual behavior

Both chats show the markers plus full message bubbles with the bots' replies and relays, so the user's DMs read as if the bots are messaging them.

Rakazo version or commit

50fa032 (self-hosted, Docker images)

Operating system

Ubuntu (DigitalOcean droplet); clients: web, macOS desktop app, iOS app

SANDBOX_PROVIDER

docker