diff --git a/frontends/tuiapp_v2.py b/frontends/tuiapp_v2.py
index 60a3be7df..e3bee9e2e 100644
--- a/frontends/tuiapp_v2.py
+++ b/frontends/tuiapp_v2.py
@@ -6164,6 +6164,12 @@ def _drain_ask_user_events(self, sess: AgentSession) -> None:
             except queue.Empty: break
         if not latest: return
         question = latest["question"]; candidates = latest["candidates"]
+        # v2 parity with tui_v3 (dfab299): when the agent didn't supply
+        # candidates, mounting an empty ChoiceList still calls .focus() on
+        # mount and steals keyboard focus from the text input. Skip the
+        # card; the 'Waiting for your answer ...' marker is already in
+        # scrollback, and the user replies via the normal input box.
+        if not candidates: return
         multi = bool(self._MULTI_RE.search(question))
         kind = "multi_choice" if multi else "choice"
         choices = [(c, c) for c in candidates] + [(FREE_TEXT_LABEL, FREE_TEXT_CHOICE)]
diff --git a/tests/test_tuiapp_v2_ask_user_drain.py b/tests/test_tuiapp_v2_ask_user_drain.py
new file mode 100644
index 000000000..792482a2b
--- /dev/null
+++ b/tests/test_tuiapp_v2_ask_user_drain.py
@@ -0,0 +1,105 @@
+"""Behavioural test for tuiapp_v2._drain_ask_user_events empty-candidates guard.
+
+Issue #615: when ask_user is called with candidates=[], _drain_ask_user_events
+used to mount an empty ChoiceList (only "Type something") and steal keyboard
+focus from the text input. Mirrors the v3 fix in dfab299. We verify:
+
+  1. With candidates=[], the drain does NOT append any ChatMessage.
+  2. With candidates=['A', 'B'], the drain still appends exactly one
+     ChatMessage (regression test — the fix must not break the happy path).
+  3. The empty-candidate event is still consumed from the queue (so a
+     stale event doesn't trigger the empty-picker bug on a later drain).
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+from queue import Queue
+
+# frontends/ has no __init__.py — add it to sys.path so sibling modules
+# (keysym, at_complete, slash_cmds, …) import cleanly.
+REPO_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
+sys.path.insert(0, os.path.join(REPO_ROOT, "frontends"))
+sys.path.insert(0, REPO_ROOT)
+
+import importlib.util
+_spec = importlib.util.spec_from_file_location(
+    "_tuiapp_v2_for_test",
+    os.path.join(REPO_ROOT, "frontends", "tuiapp_v2.py"),
+)
+t2 = importlib.util.module_from_spec(_spec)
+sys.modules["_tuiapp_v2_for_test"] = t2  # dataclass introspects sys.modules
+_spec.loader.exec_module(t2)  # type: ignore[union-attr]
+
+
+class _StubSess:
+    """Minimal stand-in for AgentSession — only the attributes the drain touches."""
+
+    def __init__(self, payload):
+        self.ask_user_events: Queue = Queue()
+        self.messages: list = []
+        self.agent_id = 1
+        self.agent = None
+        if payload is not None:
+            self.ask_user_events.put(payload)
+
+
+class _StubApp:
+    """Just enough surface for _drain_ask_user_events — it never calls .query_one."""
+
+    current_id = 1
+
+    def _refresh_messages(self):
+        # If called with an empty pick, this is the symptom of the bug.
+        raise AssertionError("_refresh_messages must not run for empty candidates")
+
+
+def _make_app_with_sess(sess):
+    """Build an instance whose _drain_ask_user_events we can call directly."""
+    app = t2.GenericAgentTUI.__new__(t2.GenericAgentTUI)
+    app.sessions = {sess.agent_id: sess}
+    app.current_id = sess.agent_id
+    # Provide the regex the method uses to detect multi-select phrasing.
+    app._MULTI_RE = t2.GenericAgentTUI._MULTI_RE
+    # No-op the side-effect that needs a real Textual screen mounted.
+    app._refresh_messages = lambda: None
+    return app
+
+
+def test_empty_candidates_no_message_appended():
+    sess = _StubSess({"question": "Pick one", "candidates": []})
+    app = _make_app_with_sess(sess)
+    t2.GenericAgentTUI._drain_ask_user_events(app, sess)
+    assert sess.messages == [], f"expected no picker, got {sess.messages!r}"
+    # The stale event must still have been consumed (queue empty).
+    assert sess.ask_user_events.qsize() == 0, "drain must still consume empty events"
+
+
+def test_non_empty_candidates_still_appends_picker():
+    sess = _StubSess({"question": "Pick one", "candidates": ["A", "B"]})
+    app = _make_app_with_sess(sess)
+    t2.GenericAgentTUI._drain_ask_user_events(app, sess)
+    assert len(sess.messages) == 1, f"expected 1 picker, got {len(sess.messages)}"
+    msg = sess.messages[0]
+    # Free-text escape hatch should still be appended.
+    labels = [label for label, _ in msg.choices]
+    assert "Type something" in labels
+    assert "A" in labels and "B" in labels
+
+
+def test_multi_select_marker_with_empty_candidates_still_skipped():
+    sess = _StubSess({"question": "[多选] tag a few", "candidates": []})
+    app = _make_app_with_sess(sess)
+    t2.GenericAgentTUI._drain_ask_user_events(app, sess)
+    assert sess.messages == [], "multi-select phrasing must not bypass empty guard"
+
+
+if __name__ == "__main__":
+    test_empty_candidates_no_message_appended()
+    print("✓ empty candidates → no picker mounted")
+    test_non_empty_candidates_still_appends_picker()
+    print("✓ non-empty candidates → picker still mounted")
+    test_multi_select_marker_with_empty_candidates_still_skipped()
+    print("✓ multi-select phrasing does not bypass empty-candidates guard")
+    print("all checks passed")
\ No newline at end of file