fix: Phase 3 line-walk fixes (3a-3d)

F3a-01 agent.py: _safe_float() prevents ValueError from non-numeric metric values
F3b-01 schedule_runner.py: fix tz-aware minus naive DateTime TypeError
F3b-02 schedule_runner.py: cron field divided by 0 returns False instead of crash
F3c-01 redis/client.py: mask credentials in REDIS_URL log output
F3d-01 main.py: cancel background tasks when primary worker lock is lost
F3d-02 main.py: strip whitespace from CORS origins after split

Co-authored-by: Cursor <cursoragent@cursor.com>

server/api/agent.py

@@ -83,6 +83,17 @@ def _threshold_for(thresholds: dict, metric: str) -> float:
     return float(thresholds.get(metric, defaults.get(metric, 80)))
 
 
+def _safe_float(value, name: str) -> Optional[float]:
+    """Convert agent-supplied metric value to float, discarding malformed values."""
+    if value is None:
+        return None
+    try:
+        return float(value)
+    except (ValueError, TypeError):
+        logger.warning(f"Agent sent non-numeric {name}: {value!r!s:.50} — skipped")
+        return None
+
+
 def _detect_alerts(system_info: dict, thresholds: dict) -> list:
     """Detect metrics exceeding alert thresholds
 
@@ -90,16 +101,16 @@ def _detect_alerts(system_info: dict, thresholds: dict) -> list:
     E.g. [("cpu", 85.3), ("mem", 92.1)]
     """
     alerts = []
-    cpu = system_info.get("cpu_usage") or system_info.get("cpu")
-    mem = system_info.get("mem_usage") or system_info.get("mem")
-    disk = system_info.get("disk_usage") or system_info.get("disk")
+    cpu = _safe_float(system_info.get("cpu_usage") or system_info.get("cpu"), "cpu")
+    mem = _safe_float(system_info.get("mem_usage") or system_info.get("mem"), "mem")
+    disk = _safe_float(system_info.get("disk_usage") or system_info.get("disk"), "disk")
 
-    if cpu is not None and float(cpu) > _threshold_for(thresholds, "cpu"):
-        alerts.append(("cpu", float(cpu)))
-    if mem is not None and float(mem) > _threshold_for(thresholds, "mem"):
-        alerts.append(("mem", float(mem)))
-    if disk is not None and float(disk) > _threshold_for(thresholds, "disk"):
-        alerts.append(("disk", float(disk)))
+    if cpu is not None and cpu > _threshold_for(thresholds, "cpu"):
+        alerts.append(("cpu", cpu))
+    if mem is not None and mem > _threshold_for(thresholds, "mem"):
+        alerts.append(("mem", mem))
+    if disk is not None and disk > _threshold_for(thresholds, "disk"):
+        alerts.append(("disk", disk))
 
     return alerts
 
@@ -114,11 +125,12 @@ def _detect_recovery(system_info: dict, prev_alerts: set, thresholds: dict) -> l
         parts = prev.split(":")
         if len(parts) != 2:
             continue
-        metric, prev_val = parts[0], float(parts[1])
-        current = system_info.get(f"{metric}_usage") or system_info.get(metric)
-
-        if current is not None and float(current) < _threshold_for(thresholds, metric):
-            recoveries.append((metric, float(current)))
+        metric = parts[0]
+        current = _safe_float(
+            system_info.get(f"{metric}_usage") or system_info.get(metric), metric
+        )
+        if current is not None and current < _threshold_for(thresholds, metric):
+            recoveries.append((metric, current))
 
     return recoveries