fix(cli): prevent stale image attachment on text paste and voice input

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

cli.py

@@ -1203,6 +1203,11 @@ def _format_image_attachment_badges(attached_images: list[Path], image_counter:
     )
 
 
+def _should_auto_attach_clipboard_image_on_paste(pasted_text: str) -> bool:
+    """Auto-attach clipboard images only for image-only paste gestures."""
+    return not pasted_text.strip()
+
+
 def _collect_query_images(query: str | None, image_arg: str | None = None) -> tuple[str, list[Path]]:
     """Collect local image attachments for single-query CLI flows."""
     message = query or ""
@@ -6282,6 +6287,9 @@ class HermesCLI:
 
             if result.get("success") and result.get("transcript", "").strip():
                 transcript = result["transcript"].strip()
+                self._attached_images.clear()
+                if hasattr(self, '_app') and self._app:
+                    self._app.invalidate()
                 self._pending_input.put(transcript)
                 submitted = True
             elif result.get("success"):
@@ -8006,8 +8014,9 @@ class HermesCLI:
             """Handle terminal paste — detect clipboard images.
 
             When the terminal supports bracketed paste, Ctrl+V / Cmd+V
-            triggers this with the pasted text.  We also check the
-            clipboard for an image on every paste event.
+            triggers this with the pasted text. We only auto-attach a
+            clipboard image for image-only/empty paste gestures so text
+            pastes and dictation do not accidentally attach stale images.
 
             Large pastes (5+ lines) are collapsed to a file reference
             placeholder while preserving any existing user text in the
@@ -8017,7 +8026,7 @@ class HermesCLI:
             # Normalise line endings — Windows \r\n and old Mac \r both become \n
             # so the 5-line collapse threshold and display are consistent.
             pasted_text = pasted_text.replace('\r\n', '\n').replace('\r', '\n')
-            if self._try_attach_clipboard_image():
+            if _should_auto_attach_clipboard_image_on_paste(pasted_text) and self._try_attach_clipboard_image():
                 event.app.invalidate()
             if pasted_text:
                 line_count = pasted_text.count('\n')

tests/tools/test_clipboard.py

@@ -35,6 +35,7 @@ from hermes_cli.clipboard import (
     _windows_has_image,
     _convert_to_png,
 )
+from cli import _should_auto_attach_clipboard_image_on_paste
 
 FAKE_PNG = b"\x89PNG\r\n\x1a\n" + b"\x00" * 100
 FAKE_BMP = b"BM" + b"\x00" * 100
@@ -919,6 +920,48 @@ class TestTryAttachClipboardImage:
         assert path.suffix == ".png"
 
 
+class TestAutoAttachClipboardImageOnPaste:
+    def test_skips_auto_attach_for_plain_text_paste(self):
+        assert _should_auto_attach_clipboard_image_on_paste("hello world") is False
+
+    def test_skips_auto_attach_for_whitespace_and_text_paste(self):
+        assert _should_auto_attach_clipboard_image_on_paste("  hello world  ") is False
+
+    def test_allows_auto_attach_for_empty_paste(self):
+        assert _should_auto_attach_clipboard_image_on_paste("") is True
+
+    def test_allows_auto_attach_for_whitespace_only_paste(self):
+        assert _should_auto_attach_clipboard_image_on_paste("   \n\t  ") is True
+
+
+class TestVoiceSubmission:
+    @pytest.fixture
+    def cli(self):
+        from cli import HermesCLI
+        cli_obj = HermesCLI.__new__(HermesCLI)
+        cli_obj._attached_images = [Path("/tmp/stale.png")]
+        cli_obj._pending_input = queue.Queue()
+        cli_obj._voice_lock = MagicMock()
+        cli_obj._voice_processing = True
+        cli_obj._voice_recording = True
+        cli_obj._voice_continuous = False
+        cli_obj._no_speech_count = 0
+        cli_obj._voice_recorder = MagicMock()
+        cli_obj._voice_recorder.stop.return_value = "/tmp/fake.wav"
+        cli_obj._app = None
+        return cli_obj
+
+    def test_voice_transcript_clears_stale_attached_images(self, cli):
+        with patch("tools.voice_mode.play_beep"):
+            with patch("tools.voice_mode.transcribe_recording", return_value={"success": True, "transcript": "hello"}):
+                with patch("os.path.isfile", return_value=False):
+                    with patch("cli._cprint"):
+                        cli._voice_stop_and_transcribe()
+
+        assert cli._attached_images == []
+        assert cli._pending_input.get_nowait() == "hello"
+
+
 # ═════════════════════════════════════════════════════════════════════════
 # Level 4: Queue routing — tuple unpacking in process_loop
 # ═════════════════════════════════════════════════════════════════════════