diff --git a/pyproject.toml b/pyproject.toml
index 8060e5d..5920541 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -12,7 +12,7 @@ requires-python = ">=3.10"
 dependencies = [
     "pyarrow>=14.0",
     "jsonschema>=4.20",
-    "quant-data-kit @ git+https://github.com/PureSaber/quant-data-kit.git@3b1684dfb1ecddd0c7df95f1fbac9b21119e9f55",
+    "quant-data-kit @ git+https://github.com/PureSaber/quant-data-kit.git@ab17ee7b8c417aa72a9707106f5455beba965b76",
 ]
 
 [project.optional-dependencies]
diff --git a/requirements.lock b/requirements.lock
index d68bab4..382e96e 100644
--- a/requirements.lock
+++ b/requirements.lock
@@ -69,7 +69,7 @@ pytz==2026.3.post1
     # via pandas
 pyyaml==6.0.3
     # via quant-data-kit
-quant-data-kit @ git+https://github.com/PureSaber/quant-data-kit.git@3b1684dfb1ecddd0c7df95f1fbac9b21119e9f55
+quant-data-kit @ git+https://github.com/PureSaber/quant-data-kit.git@ab17ee7b8c417aa72a9707106f5455beba965b76
     # via quant-execution (pyproject.toml)
 referencing==0.37.0
     # via
diff --git a/src/quant_execution/derivative_replay.py b/src/quant_execution/derivative_replay.py
index 4847127..2f6959b 100644
--- a/src/quant_execution/derivative_replay.py
+++ b/src/quant_execution/derivative_replay.py
@@ -22,6 +22,7 @@ def replay(bundle, config, decide):
     Fees and adverse tick slippage apply to every fill. Daily-bar fills are a
     declared research assumption, not bid/ask or actual execution evidence.
     """
+    bundle.require_mode("replay")
     groups = defaultdict(dict)
     for q in bundle.quotes:
         groups[q.session][q.instrument_id] = q
diff --git a/tests/test_derivative_replay.py b/tests/test_derivative_replay.py
index a30b2fd..bdcc245 100644
--- a/tests/test_derivative_replay.py
+++ b/tests/test_derivative_replay.py
@@ -9,6 +9,22 @@
 from quant_execution.derivatives import DerivativeAccount
 
 
+def test_analysis_only_bundle_cannot_enter_account_replay(tmp_path):
+    demo = write_demo(tmp_path / "demo", "future")
+    bundle = write_bundle(
+        tmp_path / "limited",
+        demo.contracts,
+        demo.quotes,
+        provider="test",
+        evidence_kind="synthetic",
+        rights_note="test",
+        limits="analysis only",
+        source={"allowed_modes": ["analysis"]},
+    )
+    with pytest.raises(ValueError, match="permits"):
+        replay(bundle, {"initial_cash": 10000}, lambda quotes, at: {})
+
+
 def test_next_bar_fills_artifact_balance_and_tamper_rejection(tmp_path):
     bundle = write_demo(tmp_path / "bundle", "future")
     config = {