fix(gateway): pass channel_prompt into voice-channel STT events; guard empty transcripts

Two hand-written fixes in the voice input path:

- _handle_voice_channel_input now resolves the bound text channel's
  channel_prompt via the adapter's _resolve_channel_prompt so voice input
  gets the same per-channel context as typed messages (fixes #50149).
- _enrich_message_with_transcription now guards success=True results whose
  transcript is empty/whitespace-only (silence, cut-off, inaudible audio):
  instead of emitting empty quotes the agent gets a clear sentinel note.
  Reimplemented against the current plain-quoted note wording; original
  concept and tests by @deacon-botdoctor in PR #41603 (fixes #41603).

gateway/run.py

@@ -15755,11 +15755,22 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
         # Use SimpleNamespace as raw_message so _get_guild_id() can extract
         # guild_id and _send_voice_reply() plays audio in the voice channel.
         from types import SimpleNamespace
+        # Resolve the bound text channel's channel_prompt so voice input gets
+        # the same per-channel context as typed messages (#50149).
+        channel_prompt: Optional[str] = None
+        resolver = getattr(adapter, "_resolve_channel_prompt", None)
+        if callable(resolver):
+            try:
+                resolved = resolver(str(text_ch_id))
+                channel_prompt = resolved if isinstance(resolved, str) else None
+            except Exception:
+                channel_prompt = None
         event = MessageEvent(
             source=source,
             text=transcript,
             message_type=MessageType.VOICE,
             raw_message=SimpleNamespace(guild_id=guild_id, guild=None),
+            channel_prompt=channel_prompt,
         )
 
         await adapter.handle_message(event)
@@ -18012,6 +18023,19 @@ class GatewayRunner(GatewayAuthorizationMixin, GatewayKanbanWatchersMixin, Gatew
                 result = await asyncio.to_thread(transcribe_audio, path)
                 if result["success"]:
                     transcript = result["transcript"]
+                    # Speech-to-text can return success=True with an empty or
+                    # whitespace-only transcript on silence, cut-off, or
+                    # inaudible audio. Emitting empty quotes ('""') makes the
+                    # agent reply to nothing and can loop, so that case gets a
+                    # clear sentinel note instead (#41603).
+                    if not (transcript or "").strip():
+                        enriched_parts.append(
+                            "[The user sent a voice message but it came through "
+                            "empty or inaudible — speech-to-text returned no "
+                            "words. Do not guess at the content; ask the user "
+                            "to resend or type it out.]"
+                        )
+                        continue
                     successful_transcripts.append(transcript)
                     # Pass the transcript through as a plain quoted line. The
                     # earlier wording ("The user sent a voice message~ Here's

tests/gateway/test_stt_config.py

@@ -181,6 +181,30 @@ async def test_enrich_message_with_transcription_handles_missing_transcription_m
     assert transcripts == []
 
 
+@pytest.mark.asyncio
+async def test_enrich_message_with_transcription_guards_empty_transcript():
+    """success=True with an empty/whitespace transcript must not emit empty
+    quotes — it gets a sentinel note and is excluded from transcripts (#41603)."""
+    from gateway.run import GatewayRunner
+
+    runner = GatewayRunner.__new__(GatewayRunner)
+    runner.config = GatewayConfig(stt_enabled=True)
+    runner._has_setup_skill = lambda: False
+
+    with patch(
+        "tools.transcription_tools.transcribe_audio",
+        return_value={"success": True, "transcript": "   \n\t", "provider": "local_command"},
+    ):
+        result, transcripts = await runner._enrich_message_with_transcription(
+            "caption",
+            ["/tmp/voice.ogg"],
+        )
+
+    assert "empty or inaudible" in result
+    assert '""' not in result
+    assert transcripts == []
+
+
 @pytest.mark.asyncio
 async def test_prepare_inbound_message_text_transcribes_queued_voice_event():
     from gateway.run import GatewayRunner

tests/gateway/test_voice_command.py

@@ -1030,6 +1030,39 @@ class TestVoiceChannelCommands:
         assert event.source.chat_id == "123"
         assert event.source.chat_type == "channel"
 
+    @pytest.mark.asyncio
+    async def test_input_resolves_channel_prompt(self, runner):
+        """Voice input must carry the bound text channel's channel_prompt (#50149)."""
+        from gateway.config import Platform
+        mock_adapter = AsyncMock()
+        mock_adapter._voice_text_channels = {111: 123}
+        mock_adapter._voice_sources = {}
+        mock_adapter._client = MagicMock()
+        mock_adapter._client.get_channel = MagicMock(return_value=AsyncMock())
+        mock_adapter.handle_message = AsyncMock()
+        mock_adapter._resolve_channel_prompt = MagicMock(return_value="Be terse in #dev.")
+        runner.adapters[Platform.DISCORD] = mock_adapter
+        await runner._handle_voice_channel_input(111, 42, "Hello from VC")
+        mock_adapter._resolve_channel_prompt.assert_called_once_with("123")
+        event = mock_adapter.handle_message.call_args[0][0]
+        assert event.channel_prompt == "Be terse in #dev."
+
+    @pytest.mark.asyncio
+    async def test_input_channel_prompt_resolver_failure_is_non_fatal(self, runner):
+        """A failing channel_prompt resolver must not block voice input."""
+        from gateway.config import Platform
+        mock_adapter = AsyncMock()
+        mock_adapter._voice_text_channels = {111: 123}
+        mock_adapter._voice_sources = {}
+        mock_adapter._client = MagicMock()
+        mock_adapter._client.get_channel = MagicMock(return_value=AsyncMock())
+        mock_adapter.handle_message = AsyncMock()
+        mock_adapter._resolve_channel_prompt = MagicMock(side_effect=RuntimeError("boom"))
+        runner.adapters[Platform.DISCORD] = mock_adapter
+        await runner._handle_voice_channel_input(111, 42, "Hello from VC")
+        event = mock_adapter.handle_message.call_args[0][0]
+        assert event.channel_prompt is None
+
     @pytest.mark.asyncio
     async def test_input_reuses_bound_source_metadata(self, runner):
         """Voice input should share the linked text channel session metadata."""