diff --git a/apps/api/src/services/decision_manager.py b/apps/api/src/services/decision_manager.py
index 10a4edc9..584aa822 100644
--- a/apps/api/src/services/decision_manager.py
+++ b/apps/api/src/services/decision_manager.py
@@ -259,6 +259,91 @@ async def _push_decision_to_telegram(
         )
 
 
+async def _fetch_metrics_snapshot(incident: Incident) -> dict:
+    """
+    ADR-071-I: 從 Prometheus 抓取與此 incident 相關的指標快照
+    失敗時靜默返回空 dict，不阻塞主流程
+
+    2026-04-11 Claude Sonnet 4.6 Asia/Taipei
+    """
+    try:
+        from src.plugins.mcp.providers.prometheus_provider import PrometheusProvider
+        prom = PrometheusProvider()
+        if not prom.enabled:
+            return {}
+
+        labels = incident.signals[0].labels if incident.signals else {}
+        alertname = labels.get("alertname", "")
+        instance  = labels.get("instance", "")
+        snapshots: dict = {}
+
+        # 根據 alertname 選擇最相關的指標
+        if alertname in ("HostHighCpuLoad", "HostOutOfMemory"):
+            if instance:
+                host = instance.split(":")[0]
+                r = await prom._instant_query(
+                    f'100 - (avg by(instance) (irate(node_cpu_seconds_total{{mode="idle",instance=~"{host}.*"}}[5m])) * 100)'
+                )
+                if r.get("status") == "success":
+                    for item in r.get("data", {}).get("result", []):
+                        snapshots["cpu_pct"] = round(float(item["value"][1]), 1)
+            r2 = await prom._instant_query(
+                f'(1 - (node_memory_MemAvailable_bytes{{instance=~"{instance}"}} / node_memory_MemTotal_bytes{{instance=~"{instance}"}})) * 100'
+                if instance else "100 * (1 - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)"
+            )
+            if r2.get("status") == "success":
+                for item in r2.get("data", {}).get("result", []):
+                    snapshots["mem_pct"] = round(float(item["value"][1]), 1)
+
+        elif alertname == "HostOutOfDiskSpace":
+            r = await prom._instant_query(
+                'max(100 - ((node_filesystem_avail_bytes{fstype!="tmpfs"} / node_filesystem_size_bytes{fstype!="tmpfs"}) * 100))'
+            )
+            if r.get("status") == "success":
+                for item in r.get("data", {}).get("result", []):
+                    snapshots["disk_pct"] = round(float(item["value"][1]), 1)
+
+        elif alertname in ("PodRestartingTooMuch", "PodCrashLoopBackOff"):
+            pod = labels.get("pod", labels.get("component", ""))
+            if pod:
+                r = await prom._instant_query(
+                    f'sum(kube_pod_container_status_restarts_total{{namespace="awoooi-prod",pod=~"{pod}.*"}})'
+                )
+                if r.get("status") == "success":
+                    for item in r.get("data", {}).get("result", []):
+                        snapshots["restart_count"] = int(float(item["value"][1]))
+
+        return snapshots
+    except Exception as _e:
+        logger.debug("metrics_snapshot_failed", incident_id=incident.incident_id, error=str(_e))
+        return {}
+
+
+def _format_metrics_delta(before: dict, after: dict) -> str:
+    """
+    ADR-071-I: 格式化指標前後對比文字
+    例：CPU 92%→23% | Mem 78%→45%
+    """
+    if not before and not after:
+        return ""
+    parts = []
+    for key, label in [("cpu_pct", "CPU"), ("mem_pct", "Mem"), ("disk_pct", "Disk")]:
+        b = before.get(key)
+        a = after.get(key)
+        if b is not None and a is not None:
+            parts.append(f"{label} {b}%→{a}%")
+        elif b is not None:
+            parts.append(f"{label} {b}% (before)")
+        elif a is not None:
+            parts.append(f"{label} {a}% (after)")
+    for key, label in [("restart_count", "Restarts")]:
+        b = before.get(key)
+        a = after.get(key)
+        if b is not None and a is not None:
+            parts.append(f"{label} {b}→{a}")
+    return " | ".join(parts)
+
+
 async def _push_auto_repair_result(
     incident: Incident,
     action: str,
@@ -270,7 +355,10 @@ async def _push_auto_repair_result(
     統帥要求: 所有狀態變更必須在原告警訊息延續，不發新訊息。
     - append_incident_update() 取 Redis tg_msg:{id} → reply 原訊息 + 換按鈕
     - 找不到 message_id 時 fallback 到 send_notification（降級）
+
+    ADR-071-I: 成功時抓 metrics_after 快照，寫入 incidents 表，並在通知中顯示前後對比
     2026-04-09 Claude Sonnet 4.6 Asia/Taipei
+    2026-04-11 Claude Sonnet 4.6: +ADR-071-I 指標快照
     """
     try:
         from src.services.telegram_gateway import get_telegram_gateway
@@ -279,10 +367,37 @@ async def _push_auto_repair_result(
         target = incident.affected_services[0] if incident.affected_services else "unknown"
         inc_id = incident.incident_id
 
+        # ADR-071-I: 抓 metrics_after（成功時）
+        metrics_delta_text = ""
         if success:
+            metrics_after = await _fetch_metrics_snapshot(incident)
+            metrics_before = getattr(incident, "metrics_before", None) or {}
+
+            # 寫入 DB（不阻塞主流程）
+            if metrics_after:
+                try:
+                    from src.db.base import get_db_context
+                    from src.db.models import Incident as IncidentORM
+                    from sqlalchemy import update as _update
+                    async with get_db_context() as db:
+                        await db.execute(
+                            _update(IncidentORM)
+                            .where(IncidentORM.incident_id == inc_id)
+                            .values(metrics_after=metrics_after)
+                        )
+                        await db.commit()
+                    logger.info("metrics_after_saved", incident_id=inc_id, metrics=metrics_after)
+                except Exception as _e:
+                    logger.warning("metrics_after_save_failed", incident_id=inc_id, error=str(_e))
+
+            metrics_delta_text = _format_metrics_delta(metrics_before, metrics_after)
+
+        if success:
+            delta_line = f"\n├ 指標: <code>{metrics_delta_text}</code>" if metrics_delta_text else ""
             status_line = (
                 f"✅ <b>自動修復完成</b>\n"
-                f"└ <code>{action[:100] if action else '已執行'}</code>"
+                f"├ <code>{action[:100] if action else '已執行'}</code>"
+                f"{delta_line}"
             )
         else:
             status_line = (
@@ -716,6 +831,24 @@ class DecisionManager:
                 risk_level=_risk,
             )
 
+            # ADR-071-I: 執行前抓 metrics_before 快照 (2026-04-11 Claude Sonnet 4.6)
+            _metrics_before = await _fetch_metrics_snapshot(incident)
+            if _metrics_before:
+                try:
+                    from src.db.base import get_db_context
+                    from src.db.models import Incident as IncidentORM
+                    from sqlalchemy import update as _sa_update
+                    async with get_db_context() as _db:
+                        await _db.execute(
+                            _sa_update(IncidentORM)
+                            .where(IncidentORM.incident_id == incident.incident_id)
+                            .values(metrics_before=_metrics_before)
+                        )
+                        await _db.commit()
+                except Exception as _mb_err:
+                    logger.debug("metrics_before_save_failed",
+                                 incident_id=incident.incident_id, error=str(_mb_err))
+
             # 執行
             executor = ApprovalExecutionService()
             await executor.execute_approved_action(approval)
diff --git a/apps/api/src/services/km_conversion_service.py b/apps/api/src/services/km_conversion_service.py
index 2bb32cd2..9d3d7531 100644
--- a/apps/api/src/services/km_conversion_service.py
+++ b/apps/api/src/services/km_conversion_service.py
@@ -209,16 +209,31 @@ class KMConversionService:
             action_type = decision_chain.get("action_type", "")
             action_command = decision_chain.get("action", "")
 
-        # 指標快照（若有）
+        # 指標快照（若有）— ADR-071-J: 使用精簡 delta 格式 (2026-04-11 Claude Sonnet 4.6)
         metrics_section = ""
         if incident.metrics_before or incident.metrics_after:
             mb = incident.metrics_before or {}
             ma = incident.metrics_after or {}
+            delta_parts = []
+            for key, label in (("cpu_pct", "CPU"), ("mem_pct", "Mem"), ("disk_pct", "Disk")):
+                if key in mb and key in ma:
+                    delta_parts.append(f"{label} {mb[key]:.0f}%→{ma[key]:.0f}%")
+                elif key in ma:
+                    delta_parts.append(f"{label} 後={ma[key]:.0f}%")
+            if "restart_count" in mb and "restart_count" in ma:
+                delta_parts.append(f"重啟 {mb['restart_count']}→{ma['restart_count']}")
+            delta_str = " | ".join(delta_parts) if delta_parts else "（無量化數據）"
+
+            # K8s 狀態（若有）
+            k8s_state = getattr(incident, "k8s_state_after", None)
+            k8s_line = f"\n- K8s 狀態: {k8s_state}" if k8s_state else ""
+
             metrics_section = (
                 f"\n## 效果驗證\n"
-                f"- 執行前: {mb}\n"
-                f"- 執行後: {ma}\n"
+                f"- 指標變化: {delta_str}\n"
                 f"- 恢復耗時: {resolution_time}\n"
+                + k8s_line
+                + "\n"
             )
 
         # 驗證結果（若有）