feat: prepare 2.2.0 with UI and execution fixes

Add dark theme, richer Markdown and readable plugin results.
Improve Android input evidence, response integrity and diagnostics.
Unify lifecycle, guidance receipts and evidence sources across clients.
Keep direct action dispatch without an added AI quota approval layer.
Align version metadata at 2.2.0 and bundle Helper 1.2.2.

Validation: 3063 backend tests before the version bump; 44 after.
Frontend: 317 tests and production build passed.

tests/unit/agents/test_flash_progress.py

@@ -72,6 +72,49 @@ def test_pc_first_and_intentional_interleaving_are_supported():
     assert progress.guard("cua-driver") is None
 
 
+def test_stage_updates_are_atomic_and_preserve_unmentioned_scope_and_order(tmp_path):
+    progress = TaskProgress()
+    resources = {"android", "cua-driver"}
+    progress.update(plan(), resources=resources, directory=tmp_path)
+    before = progress.value.model_dump()
+    saved = (tmp_path / "notes/task_progress.md").read_bytes()
+    with pytest.raises(ValidationError, match="完成或阻塞阶段必须记录"):
+        progress.update(
+            {"reason": "无证据不能完成", "stage_updates": [{"id": "phone", "status": "done"}]},
+            resources=resources,
+            directory=tmp_path,
+        )
+    assert progress.value.model_dump() == before
+    assert (tmp_path / "notes/task_progress.md").read_bytes() == saved
+    progress.update(
+        {
+            "reason": "手机已检查完",
+            "stage_updates": [
+                {"id": "phone", "status": "done", "evidence": "实际步骤记录 real-step"}
+            ],
+        },
+        resources=resources,
+        directory=tmp_path,
+    )
+    assert progress.value.stages[1].model_dump() == before["stages"][1]
+    assert [s.id for s in progress.value.stages] == ["phone", "pc"]
+    assert progress.value.stages[0].scope == before["stages"][0]["scope"]
+    progress.update(
+        {"reason": "开始电脑检查", "stage_updates": [{"id": "pc", "status": "active"}]},
+        resources=resources,
+    )
+    assert progress.guard("cua-driver") is None
+    with pytest.raises(ValueError, match="未知 id"):
+        progress.update(
+            {"reason": "错拼不能覆盖", "stage_updates": [{"id": "typo", "status": "done"}]},
+            resources=resources,
+        )
+    with pytest.raises(ValidationError, match="只能提供一种"):
+        progress.update(
+            {**plan(), "stage_updates": [{"id": "pc", "status": "active"}]}, resources=resources
+        )
+
+
 @pytest.mark.asyncio
 async def test_flash_preflight_does_not_authorize_pc_mutation(monkeypatch):
     plugins = SimpleNamespace(