diff --git a/metrics/__init__.py b/metrics/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/metrics/len_var/README.md b/metrics/len_var/README.md
new file mode 100644
index 00000000..5b77b8cc
--- /dev/null
+++ b/metrics/len_var/README.md
@@ -0,0 +1,8 @@
+# Length-Variance (len_var) Scorer
+
+| Key | Value |
+|-----|-------|
+| **Purpose** | Detects internal length fluctuation in a single answer |
+| **Formula** | σ(window_lengths) / μ(window_lengths) |
+| **Threshold (tentative)** | <0.20 : ✅ / 0.20-0.39 : ⚠️ / ≥0.40 : 🔥 |
+| **Dependencies** | None (pure Python stdlib) |
diff --git a/metrics/len_var/__init__.py b/metrics/len_var/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/metrics/len_var/scorer.py b/metrics/len_var/scorer.py
new file mode 100644
index 00000000..724b774d
--- /dev/null
+++ b/metrics/len_var/scorer.py
@@ -0,0 +1,17 @@
+import statistics as st
+from typing import List
+
+def _window_lengths(tokens: List[str], window: int = 5) -> List[int]:
+    return [len(" ".join(tokens[i:i+window]))
+            for i in range(len(tokens) - window + 1)]
+
+def score(prediction: str, reference: str = "", window: int = 5) -> float:
+    """
+    Length-variance score (0 = 安定, 1 ≒ 大揺れ)
+    定義: sliding-window の文字長 標準偏差 ÷ 平均長
+    """
+    toks = prediction.split()
+    if len(toks) < window:
+        return 0.0
+    lens = _window_lengths(toks, window)
+    return st.pstdev(lens) / max(st.mean(lens), 1)
diff --git a/metrics/len_var/tests/test_len_var.py b/metrics/len_var/tests/test_len_var.py
new file mode 100644
index 00000000..c51d2da6
--- /dev/null
+++ b/metrics/len_var/tests/test_len_var.py
@@ -0,0 +1,9 @@
+from metrics.len_var.scorer import score
+
+def test_constant_length():
+    s = "word " * 10
+    assert score(s) < 0.05
+
+def test_varied_length():
+    s = "a a a a a " + "this sentence is considerably longer "
+    assert score(s) > 0.2
diff --git a/types.py b/project_types.py
similarity index 100%
rename from types.py
rename to project_types.py
diff --git a/queue_stub.py b/queue_stub.py
new file mode 100644
index 00000000..086a1c13
--- /dev/null
+++ b/queue_stub.py
@@ -0,0 +1,19 @@
+import threading, hashlib
+
+class DuplicateGuardQueue:
+    def __init__(self):
+        self._lock = threading.Lock()
+        self._seen = set()
+        self._queue = []
+
+    def put_unique(self, msg: str):
+        h = hashlib.sha256(msg.encode()).hexdigest()
+        with self._lock:
+            if h in self._seen:
+                return  # duplicate suppressed
+            self._seen.add(h)
+            self._queue.append(msg)
+
+    # helper for tests
+    def count(self, msg: str):
+        return self._queue.count(msg)
diff --git a/test_duplicate_send.py b/test_duplicate_send.py
new file mode 100644
index 00000000..edc8c1e2
--- /dev/null
+++ b/test_duplicate_send.py
@@ -0,0 +1,13 @@
+import threading
+from queue_stub import DuplicateGuardQueue
+
+def worker(q, msg):
+    q.put_unique(msg)
+
+def test_duplicate_blocked():
+    q = DuplicateGuardQueue()
+    text = "Saturated PoR answer"
+    threads = [threading.Thread(target=worker, args=(q, text)) for _ in range(5)]
+    for t in threads: t.start()
+    for t in threads: t.join()
+    assert q.count(text) == 1