fix(cua): SoM click must focus the input — the core of the type-stall

src/mantis_agent/gym/runner.py

@@ -1739,6 +1739,18 @@ def _save_frame(img: Any, step: int) -> None:
                         )
 
                 gym_result = self.env.step(action)
+                # type-stall fix: a SoM-dispatched click swaps the action to a
+                # no-op WAIT, so env.step's CLICK-branch focus probe never runs.
+                # If the SoM click focused a text field (cdp_click_at_point did
+                # el.focus()), surface focused_input here so the brain is told
+                # to type next instead of stalling on "no visible change".
+                if pending_executor_backend == "som":
+                    som_diag = getattr(self.env, "_last_som_diag", None)
+                    if isinstance(som_diag, dict) and som_diag.get("focused"):
+                        gym_result.info["focused_input"] = {
+                            "focused_via_click": True,
+                            "value": "",
+                        }
                 action_history.append(action)
                 for post_action in post_actions:
                     logger.warning(

src/mantis_agent/gym/xdotool_env.py

@@ -1124,9 +1124,22 @@ def cdp_click_at_point(self, x: int, y: int) -> bool:
             "(e.className && typeof e.className === 'string' ? '.' + e.className.split(' ').slice(0,2).join('.') : '')) : null;"
             "const text = e => e ? (e.innerText || e.textContent || '').trim().slice(0, 40) : '';"
             "let ok = false;"
-            "if (elv) { try { elv.click(); ok = true; } catch (e) { ok = false; } }"
+            "let focused = false;"
+            # type-stall fix: a real mouse click focuses an input on mousedown,
+            # but ``elv.click()`` only fires the click event — it does NOT focus.
+            # So SoM clicks on a search/login/composer field left it UNFOCUSED;
+            # the brain then had no field to type into and stalled (it never
+            # reached a type()). Focus focusable targets first so the next
+            # TYPE lands. Mirrors native click ordering (focus → click).
+            "if (elv) { try {"
+            "  const tagU = (elv.tagName || '').toUpperCase();"
+            "  if ((tagU === 'INPUT' || tagU === 'TEXTAREA' || tagU === 'SELECT' || elv.isContentEditable) && typeof elv.focus === 'function') {"
+            "    try { elv.focus(); focused = (document.activeElement === elv); } catch (e) {}"
+            "  }"
+            "  elv.click(); ok = true;"
+            "} catch (e) { ok = false; } }"
             "return {"
-            "ok: ok,"
+            "ok: ok, focused: focused,"
             "outerHeight: oh, innerHeight: ih, chromeH: chromeH,"
             "vx: vx, vy: vy,"
             "elv_tag: tag(elv), elv_text: text(elv),"
@@ -1157,6 +1170,9 @@ def cdp_click_at_point(self, x: int, y: int) -> bool:
             "els_tag": str(result.get("els_tag") or ""),
             "els_text": str(result.get("els_text") or ""),
             "ok": bool(result.get("ok")),
+            # type-stall fix: did the SoM click focus a text field? Read by
+            # the runner to surface focused_input → "type now" feedback.
+            "focused": bool(result.get("focused")),
         }
         # One-line WARNING log so the SoM click's coordinate translation
         # is visible in Modal's INFO-suppressed capture without an

tests/test_cua_input_focus_feedback.py

@@ -72,3 +72,33 @@ def _boom(self):
     monkeypatch.setattr(XdotoolGymEnv, "_active_field_probe", _boom)
     result = click_env.step(_click())  # must not raise
     assert "focused_input" not in result.info
+
+
+def test_som_click_focuses_input_and_reports_focused(monkeypatch: pytest.MonkeyPatch) -> None:
+    """cdp_click_at_point must el.focus() a focusable target and surface
+    ``focused`` on _last_som_diag — the type-stall fix for the SoM path, where
+    el.click() alone never focused the input."""
+    env = XdotoolGymEnv.__new__(XdotoolGymEnv)
+    env.som_diagnostic = False
+    captured: dict = {}
+
+    def _fake_eval(self, js):  # noqa: ANN001
+        captured["js"] = js
+        return {
+            "ok": True, "focused": True,
+            "elv_tag": "INPUT", "elv_text": "",
+            "els_tag": "DIV", "els_text": "",
+            "outerHeight": 800, "innerHeight": 720, "chromeH": 80,
+            "vx": 270, "vy": 42,
+        }
+
+    monkeypatch.setattr(XdotoolGymEnv, "cdp_evaluate", _fake_eval)
+
+    ok = env.cdp_click_at_point(270, 122)
+
+    assert ok is True
+    # the JS payload focuses focusable elements before clicking
+    assert "focus" in captured["js"]
+    assert ".focus()" in captured["js"]
+    # the focus result is surfaced for the runner to read
+    assert env._last_som_diag["focused"] is True