fix: distinguish lifecycle health from runtime failures

ops/quant-monitor/AGENTS.md

@@ -21,7 +21,7 @@ VPS Codex 定时监控（`codex-quant.timer` 每 30 分钟）+ 收盘简报（`c
 
 1. `sync_strategy_repos.sh` — `git pull` 四策略仓 + QPK
 2. `health_cycle.py` — `build_dashboard` + `run_drift_detection`
-3. `overall_score < 60` → Telegram 量化哨兵
+3. lifecycle `overall_score < 60` → Telegram 量化哨兵（研究/回测健康，不代表平台执行故障）
 4. drift ≥ 0.50（~2σ）→ `create_issues_for_domain` 开 Issue（不 @）
 5. drift ≥ 0.75（~3σ）→ Telegram + Issue **@owner**
 

ops/quant-monitor/scripts/health_cycle.py

@@ -161,11 +161,19 @@ def _strategy_health_alert(row: dict[str, Any]) -> tuple[str, str] | None:
     profile = str(row.get("strategy_profile") or "?")
     domain = str(row.get("domain") or "?")
     return (
-        f"[{domain}] {profile}: health_score={score:.1f}",
+        f"[{domain}] {profile}: lifecycle_health_score={score:.1f}",
         f"strategy_health_below_{SCORE_ALERT:g}:{domain}:{profile}",
     )
 
 
+def _build_alert_body(lines: list[str]) -> str:
+    return (
+        "🚨 quant-monitor strategy_lifecycle\n"
+        "• scope: research/backtest lifecycle health; not platform runtime execution\n"
+        + "\n".join(f"• {line}" for line in lines)
+    )
+
+
 def _alert_state_path(root: Path) -> Path:
     return root / _ALERT_STATE_RELATIVE_PATH
 
@@ -380,7 +388,7 @@ def main() -> int:
     telegram_sent = False
     duplicate_alert_suppressed = False
     if notify_lines:
-        body = "🚨 quant-monitor health_cycle\n" + "\n".join(f"• {line}" for line in notify_lines)
+        body = _build_alert_body(notify_lines)
         fingerprint = _alert_fingerprint(alert_identities)
         duplicate_alert_suppressed = _is_duplicate_alert(root, fingerprint)
         if not duplicate_alert_suppressed:

ops/quant-monitor/tests/test_monitor_fail_closed.py

@@ -202,6 +202,22 @@ def test_health_cycle_score_changes_keep_the_same_incident_fingerprint(self) ->
             HEALTH_CYCLE._alert_fingerprint([second_identity]),
         )
 
+    def test_health_cycle_labels_strategy_scores_as_lifecycle_not_runtime(self) -> None:
+        line, _identity = HEALTH_CYCLE._strategy_health_alert(
+            {
+                "domain": "us_equity",
+                "strategy_profile": "example",
+                "overall_score": 59.9,
+            }
+        )
+
+        body = HEALTH_CYCLE._build_alert_body([line])
+
+        self.assertIn("quant-monitor strategy_lifecycle", body)
+        self.assertIn("research/backtest lifecycle health", body)
+        self.assertIn("not platform runtime execution", body)
+        self.assertIn("lifecycle_health_score=59.9", body)
+
     def test_health_cycle_non_object_alert_state_is_a_cache_miss(self) -> None:
         with tempfile.TemporaryDirectory() as tmp:
             root = Path(tmp)