Problem

The Slack typing indicator ("is thinking...") persists indefinitely after the LLM completes a turn. It never clears.

Observed Behavior

From a customer agent thread, after the LLM finishes its response:

• The typing indicator at the bottom stays on permanently
• It only briefly reappears when the user sends a new message (at which point it gets cleared again)
• Slack's assistant.threads.setStatus is persistent (unlike Discord's auto-expiring typing)

Root Cause

In src/Netclaw.Channels.Slack/SlackThreadBindingActor.cs, the TurnCompleted message handler in HandleOutputAsync() (line ~1121) handles:

• Cursor advancement
• Reminder delivery observers
• Fallback posting
• Pending approval cleanup
• Resetting _postedThisTurn / _uploadedFileThisTurn flags

But it never calls QueueProcessingIndicatorClearIfActive() before breaking.

Code Evidence

The QueueProcessingIndicatorClearIfActive() method exists and is called in other places:

• Line 139: PostStop() (graceful shutdown)
• Line 427: When enqueueing a new inbound message (aborts previous turn)
• Line 1024: ReinitializePipelineAsync()

But it's absent from the TurnCompleted handler — the normal completion path.

Contrast with Discord

Discord uses TriggerTypingAsync() which is a transient pulse (auto-expires after ~10 seconds). The bug is masked because Discord cleans up after itself. Slack's SetStatus() is persistent and requires explicit clearing.

Suggested Fix

Add QueueProcessingIndicatorClearIfActive() before the break in the TurnCompleted case:

case TurnCompleted completed:
    // ... existing code ...
    _pendingApprovalRequests.Clear();

    // Clear the processing indicator when the turn completes
    QueueProcessingIndicatorClearIfActive();

    break;

Files to Change

• src/Netclaw.Channels.Slack/SlackThreadBindingActor.cs (line ~1170)