Grok Bot routine marks succeeded but never posts a chat bubble

Support Bug Reports
, ,
1

Describe the Bug

I hit a Grok Bot routine bug. A scheduled routine can mark succeeded and still never show a message in the agent chat. The work ran. Status is ok. I get no chat bubble. From my side it looks like the routine never ran.

The prompt already said the agent has to message me in that chat, even if there’s nothing new, and not finish a run with no message.

What I saw:

  • The scheduler fired on time
  • Status was ok
  • Internal notes were updated
  • No chat bubble. The agent later confirmed it never sent one. There was no visible routine wake in the chat. A brief only showed up later as a hidden completion payload, not a visible message.

Other due routines that morning also fired and marked ok. The clocks work. This is a delivery bug, not a missed cron.

A stay-quiet-if-nothing routine plus a hidden completion looks the same as a dead routine. Don’t treat hidden completion as something I actually saw.

Please fix:

  1. If the prompt says I should get a message, the routine wake has to show in the agent chat, not only as hidden completion.
  2. status=ok should mean that message was sent when the prompt requires one.
  3. Reading run status alone can’t tell a silent-success from a real pass.

I need routines to actually talk to me.

Steps to Reproduce

Enable several routines and they may fail

Expected Behavior

the routines should not be failing silently

Operating System

MacOS

Version Information

Grok Bot Version 0.29.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

6

Hey @kenners22, thanks for the detailed report.

This is a known issue we’re already tracking, and I’ve added your report to it. The routine does run, but its report is currently held in the background instead of being posted to the chat right away, which is why the run shows ok with no message.

In the meantime, sending the agent any message will surface the held update, since pending results are delivered on the agent’s next turn.

We’re working on making routine results post to the chat reliably, and on making run status reflect whether a message was actually delivered. We’ll follow up here when there’s an update!