feat(market): verified market-data snapshot to ground numeric claims

The market analyst could confabulate exact figures — citing a Bollinger
band or a "historically validated bounce" the data doesn't support (#830).
Add a deterministic get_verified_market_snapshot tool (latest OHLCV row,
common indicators, recent closes) the analyst must consult and treat as
the source of truth for any exact price/indicator claim, and instruct it
not to assert historical validation or support bounces without tool-backed
dates and prices.

#830

tests/test_market_data_validator.py

@@ -0,0 +1,76 @@
+"""Tests for the deterministic market-data verification snapshot (#830/#881)."""
+
+from __future__ import annotations
+
+import pandas as pd
+import pytest
+
+import tradingagents.dataflows.market_data_validator as validator
+
+
+def _sample_ohlcv() -> pd.DataFrame:
+    dates = pd.bdate_range("2026-04-01", "2026-05-20")
+    closes = [100 + i for i in range(len(dates))]
+    return pd.DataFrame({
+        "Date": dates,
+        "Open": [c - 0.5 for c in closes],
+        "High": [c + 1.0 for c in closes],
+        "Low": [c - 1.0 for c in closes],
+        "Close": closes,
+        "Volume": [1_000_000 + i for i in range(len(dates))],
+    })
+
+
+@pytest.mark.unit
+class TestVerifiedSnapshot:
+    def test_excludes_future_rows(self, monkeypatch):
+        data = pd.concat([
+            _sample_ohlcv(),
+            pd.DataFrame({"Date": [pd.Timestamp("2026-06-01")], "Open": [999.0],
+                          "High": [999.0], "Low": [999.0], "Close": [999.0], "Volume": [999]}),
+        ], ignore_index=True)
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: data)
+
+        snap = validator.build_verified_market_snapshot("COF", "2026-05-13")
+        assert "Verified market data snapshot for COF" in snap
+        assert "Requested analysis date: 2026-05-13" in snap
+        assert "Latest trading row used: 2026-05-13" in snap
+        assert "999.00" not in snap          # future row excluded
+        assert "boll_lb" in snap             # indicators present
+
+    def test_uses_previous_trading_day_when_date_is_weekend(self, monkeypatch):
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: _sample_ohlcv())
+        # 2026-05-16 is a Saturday; latest row should be Fri 2026-05-15
+        snap = validator.build_verified_market_snapshot("COF", "2026-05-16")
+        assert "Latest trading row used: 2026-05-15" in snap
+        assert "Recent verified closes" in snap
+
+    def test_raises_when_no_rows_on_or_before_date(self, monkeypatch):
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: _sample_ohlcv())
+        with pytest.raises(ValueError):
+            validator.build_verified_market_snapshot("COF", "2020-01-01")
+
+    def test_raises_on_empty_data(self, monkeypatch):
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: pd.DataFrame())
+        with pytest.raises(ValueError):
+            validator.build_verified_market_snapshot("COF", "2026-05-13")
+
+    def test_look_back_window_capped_at_30(self, monkeypatch):
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: _sample_ohlcv())
+        snap = validator.build_verified_market_snapshot("COF", "2026-05-20", look_back_days=999)
+        # last-N closes table has at most 30 data rows
+        close_rows = [ln for ln in snap.splitlines() if ln.startswith("| 2026-")]
+        assert 0 < len(close_rows) <= 30
+
+
+@pytest.mark.unit
+class TestTool:
+    def test_tool_delegates_to_builder(self, monkeypatch):
+        from tradingagents.agents.utils.market_data_validation_tools import (
+            get_verified_market_snapshot,
+        )
+        monkeypatch.setattr(validator, "load_ohlcv", lambda s, d: _sample_ohlcv())
+        out = get_verified_market_snapshot.invoke(
+            {"symbol": "COF", "curr_date": "2026-05-20"}
+        )
+        assert "Verified market data snapshot for COF" in out