diff --git a/backend/internal/handler/gateway_handler.go b/backend/internal/handler/gateway_handler.go
index c28ee846..c2b6bf09 100644
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -971,6 +971,10 @@ func (h *GatewayHandler) handleFailoverExhausted(c *gin.Context, failoverErr *se
 				msg = *rule.CustomMessage
 			}
 
+			if rule.SkipMonitoring {
+				c.Set(service.OpsSkipPassthroughKey, true)
+			}
+
 			h.handleStreamingAwareError(c, respCode, "upstream_error", msg, streamStarted)
 			return
 		}
diff --git a/backend/internal/handler/gemini_v1beta_handler.go b/backend/internal/handler/gemini_v1beta_handler.go
index f8fb0dcb..3d25505b 100644
--- a/backend/internal/handler/gemini_v1beta_handler.go
+++ b/backend/internal/handler/gemini_v1beta_handler.go
@@ -554,6 +554,10 @@ func (h *GatewayHandler) handleGeminiFailoverExhausted(c *gin.Context, failoverE
 				msg = *rule.CustomMessage
 			}
 
+			if rule.SkipMonitoring {
+				c.Set(service.OpsSkipPassthroughKey, true)
+			}
+
 			googleError(c, respCode, msg)
 			return
 		}
diff --git a/backend/internal/handler/openai_gateway_handler.go b/backend/internal/handler/openai_gateway_handler.go
index 835297b8..c08a8b0e 100644
--- a/backend/internal/handler/openai_gateway_handler.go
+++ b/backend/internal/handler/openai_gateway_handler.go
@@ -354,6 +354,10 @@ func (h *OpenAIGatewayHandler) handleFailoverExhausted(c *gin.Context, failoverE
 				msg = *rule.CustomMessage
 			}
 
+			if rule.SkipMonitoring {
+				c.Set(service.OpsSkipPassthroughKey, true)
+			}
+
 			h.handleStreamingAwareError(c, respCode, "upstream_error", msg, streamStarted)
 			return
 		}
diff --git a/backend/internal/handler/ops_error_logger.go b/backend/internal/handler/ops_error_logger.go
index 2dbf60ad..cb62ceae 100644
--- a/backend/internal/handler/ops_error_logger.go
+++ b/backend/internal/handler/ops_error_logger.go
@@ -537,6 +537,13 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 			// Store request headers/body only when an upstream error occurred to keep overhead minimal.
 			entry.RequestHeadersJSON = extractOpsRetryRequestHeaders(c)
 
+			// Skip logging if a passthrough rule with skip_monitoring=true matched.
+			if v, ok := c.Get(service.OpsSkipPassthroughKey); ok {
+				if skip, _ := v.(bool); skip {
+					return
+				}
+			}
+
 			enqueueOpsErrorLog(ops, entry, requestBody)
 			return
 		}
diff --git a/backend/internal/service/antigravity_gateway_service.go b/backend/internal/service/antigravity_gateway_service.go
index a517f243..a110f4e0 100644
--- a/backend/internal/service/antigravity_gateway_service.go
+++ b/backend/internal/service/antigravity_gateway_service.go
@@ -3251,6 +3251,21 @@ func (s *AntigravityGatewayService) writeMappedClaudeError(c *gin.Context, accou
 		log.Printf("[antigravity-Forward] upstream_error status=%d body=%s", upstreamStatus, truncateForLog(body, maxBytes))
 	}
 
+	// 检查错误透传规则
+	if ptStatus, ptErrType, ptErrMsg, matched := applyErrorPassthroughRule(
+		c, account.Platform, upstreamStatus, body,
+		0, "", "",
+	); matched {
+		c.JSON(ptStatus, gin.H{
+			"type":  "error",
+			"error": gin.H{"type": ptErrType, "message": ptErrMsg},
+		})
+		if upstreamMsg == "" {
+			return fmt.Errorf("upstream error: %d", upstreamStatus)
+		}
+		return fmt.Errorf("upstream error: %d message=%s", upstreamStatus, upstreamMsg)
+	}
+
 	var statusCode int
 	var errType, errMsg string
 
diff --git a/backend/internal/service/ops_upstream_context.go b/backend/internal/service/ops_upstream_context.go
index 65048147..3514df79 100644
--- a/backend/internal/service/ops_upstream_context.go
+++ b/backend/internal/service/ops_upstream_context.go
@@ -107,6 +107,37 @@ func appendOpsUpstreamError(c *gin.Context, ev OpsUpstreamErrorEvent) {
 	evCopy := ev
 	existing = append(existing, &evCopy)
 	c.Set(OpsUpstreamErrorsKey, existing)
+
+	checkSkipMonitoringForUpstreamEvent(c, &evCopy)
+}
+
+// checkSkipMonitoringForUpstreamEvent checks whether the upstream error event
+// matches a passthrough rule with skip_monitoring=true and, if so, sets the
+// OpsSkipPassthroughKey on the context.  This ensures intermediate retry /
+// failover errors (which never go through the final applyErrorPassthroughRule
+// path) can still suppress ops_error_logs recording.
+func checkSkipMonitoringForUpstreamEvent(c *gin.Context, ev *OpsUpstreamErrorEvent) {
+	if ev.UpstreamStatusCode == 0 {
+		return
+	}
+
+	svc := getBoundErrorPassthroughService(c)
+	if svc == nil {
+		return
+	}
+
+	// Use the best available body representation for keyword matching.
+	// Even when body is empty, MatchRule can still match rules that only
+	// specify ErrorCodes (no Keywords), so we always call it.
+	body := ev.Detail
+	if body == "" {
+		body = ev.Message
+	}
+
+	rule := svc.MatchRule(ev.Platform, ev.UpstreamStatusCode, []byte(body))
+	if rule != nil && rule.SkipMonitoring {
+		c.Set(OpsSkipPassthroughKey, true)
+	}
 }
 
 func marshalOpsUpstreamErrors(events []*OpsUpstreamErrorEvent) *string {