test(review): 把 5 处既有瑕疵锁成 *_known_gap 用例

让既有瑕疵可见，而不是被当成正确行为；改 1~3（确定性检查前置/结论结构化/
证据交叉校验）时可据此判断行为是否被改变。

- bbox_to_window 含非数值 → ValueError（上游只判长度）
- summarize / extract_issues 硬下标 f['frame'] → KeyError
- annotate 的 created 无去重（同 handle 产生重复批注）
- extract_issues 单行 ```json[...]``` 围栏剥离失败 → []
- result_json 文本 'null' → None（与「无 text 返回 default」语义易混）

tests/test_annotator.py

@@ -120,6 +120,21 @@ def test_annotate_handles_multiple_created_entities():
     assert created == ["N1", "N2"]
 
 
+def test_annotate_created_not_deduped_known_gap():
+    """已知瑕疵：同一 handle 出现两次会创建两条批注（created 无去重，仅 highlighted 去重）。"""
+    ctx = make_ctx(client=build_client())
+    created, highlighted = annotate(
+        ctx,
+        [
+            {"handle": "A", "severity": "high", "message": "第一次"},
+            {"handle": "A", "severity": "high", "message": "第二次"},
+        ],
+        log=lambda *a: None,
+    )
+    assert created == ["NEW1", "NEW1"]  # 重复批注（期望去重）
+    assert highlighted == ["A"]  # 高亮已去重
+
+
 def test_annotate_propagates_protocol_error_on_detail():
     ctx = make_ctx(client=FakeMcpClient(responses={"get_entity_detail": McpError(-32602, "bad")}))
     # annotate 只兜 add_leader/highlight 的错；detail 的协议错误会向上冒泡（现状行为）

tests/test_evidence.py

@@ -1,5 +1,7 @@
 """`dwg_review.phases.evidence` 单元测试。"""
 
+import pytest
+
 from dwg_review.phases.evidence import bbox_to_window, collect_frame
 from tests.fakes import FakeMcpClient, frame, make_ctx, ok_image, ok_text
 
@@ -13,6 +15,12 @@ def test_bbox_to_window_mapping():
     }
 
 
+def test_bbox_to_window_non_numeric_known_gap():
+    """已知瑕疵：4 元素但含非数值的 bbox 会抛 ValueError，上游 collect_frame 只判长度。"""
+    with pytest.raises(ValueError):
+        bbox_to_window([0, "a", 10, 10])
+
+
 def test_collect_frame_returns_none_for_missing_bbox():
     ctx = make_ctx(client=FakeMcpClient())
     assert collect_frame(ctx, 1, {}) is None

tests/test_mcp_client.py

@@ -63,6 +63,11 @@ def test_result_json_returns_default_when_no_text():
     assert result_json(None, "D") == "D"
 
 
+def test_result_json_null_text_known_gap():
+    """已知瑕疵：文本 'null' 是合法 JSON，解析为 None，与「无 text 返回 default」语义易混。"""
+    assert result_json(ok_text("null"), "D") is None
+
+
 def test_result_texts_tolerates_malformed_response():
     assert result_texts(None) == []
     assert result_texts({}) == []

tests/test_report.py

@@ -2,6 +2,8 @@
 
 import json
 
+import pytest
+
 from dwg_review.phases.report import extract_issues, save_report, summarize
 from tests.fakes import FakeLlmClient, make_ctx
 
@@ -44,6 +46,13 @@ def test_summarize_falls_back_to_raw_input_on_error():
     assert any("汇总失败" in line for line in logs)
 
 
+def test_summarize_missing_frame_key_known_gap():
+    """已知瑕疵：summarize 用硬下标 f['frame']，findings 缺该键直接 KeyError。"""
+    ctx = make_ctx(llm=FakeLlmClient(text="x"))
+    with pytest.raises(KeyError):
+        summarize(ctx, [{"finding": "缺 frame 键"}], log=lambda *a: None)
+
+
 # ---------------------------------------------------------------------------
 #  extract_issues
 # ---------------------------------------------------------------------------
@@ -85,6 +94,20 @@ def test_extract_issues_returns_empty_when_not_a_list():
     assert extract_issues(ctx, findings_fixture(), log=lambda *a: None) == []
 
 
+def test_extract_issues_single_line_fence_known_gap():
+    """已知瑕疵：围栏剥离依赖换行，单行 ```json[...]``` 会残留 json 前缀 → 解析失败。"""
+    llm = FakeLlmClient(text_script=['```json[{"frame": 1, "message": "z"}]```'])
+    ctx = make_ctx(llm=llm)
+    assert extract_issues(ctx, findings_fixture(), log=lambda *a: None) == []
+
+
+def test_extract_issues_missing_frame_key_known_gap():
+    """已知瑕疵：extract_issues 同样用硬下标 f['frame']，findings 缺该键直接 KeyError。"""
+    ctx = make_ctx(llm=FakeLlmClient(text="[]"))
+    with pytest.raises(KeyError):
+        extract_issues(ctx, [{"finding": "缺 frame 键"}], log=lambda *a: None)
+
+
 # ---------------------------------------------------------------------------
 #  save_report
 # ---------------------------------------------------------------------------