Now all channels can duplicate outbound replies when the agent proactively sends visible content via message(action=send).

Summary

If the agent successfully sends an image/file/text via message.send(...), OpenClaw should suppress the normal main reply / delivery-mirror path for that turn.

Right now both paths can fire:

1. The agent calls message.send(...)
2. The attachment / visible content is delivered successfully
3. The system then also runs a follow-up layer with:
   • provider=openclaw
   • model=delivery-mirror
4. A second extra message appears, often just mirroring text like the filename (for example cat-fresh.png)

Observed behavior

This causes duplicate sends across channels:

• first: attachment/image/file is sent by message.send
• then: an extra text message is posted afterward by delivery-mirror

From session history, the chain looks like:

1. message.send(...) succeeds
2. the system later emits another outbound message through provider=openclaw, model=delivery-mirror

Expected behavior

If the current assistant turn has already successfully delivered user-visible content via message(action=send), then OpenClaw should fully suppress the normal mirrored/main reply path for that same turn.

Suggested fix:

• mark the turn as externally_delivered = true after successful message.send
• skip delivery-mirror for that turn
• or discard any non-NO_REPLY main reply output once external delivery already happened

One-line version

message.send and delivery-mirror are currently both active for the same turn, so proactive sends get mirrored and users see duplicate messages.