diff --git a/docs/LOGBOOK.md b/docs/LOGBOOK.md
index 9f19a855..0ac57b18 100644
--- a/docs/LOGBOOK.md
+++ b/docs/LOGBOOK.md
@@ -158,6 +158,29 @@
 - Runbook / 操作清理：`40% -> 70%`；已補只讀診斷腳本與現況 runbook，尚待 111 恢復後補 recovery success evidence。
 - 完整 AI 自動化飛輪：維持 `67%`；本輪是 inference fallback 可觀測性與技術債清理，沒有新增 auto-execution 或 KM writeback。
 
+## 2026-06-04｜ArgoCD Application Health 設定 GitOps 化
+
+**背景**：上一階段 live 修復 `awoooi-prod` ArgoCD top-level `Degraded` 時，確認根因包含 `argocd-cm` 舊的全域 `/status` ignore 與 `argocd-cmd-params-cm` 未啟用 child resource health persistence。當時 live 已修好，但 repo 沒有 Argo configmap source-of-truth，重裝或手動漂移後可能再次退回 opaque degraded。
+
+**本次調整**：
+- 新增 `scripts/ops/apply-argocd-health-config.sh`，以 idempotent 方式維持 Argo health 設定：
+  - 若 `argocd-cm.data.resource.customizations.ignoreResourceUpdates.all` 存在，移除舊的全域 `/status` ignore。
+  - 若 `argocd-cmd-params-cm.data.controller.resource.health.persist` 不是 `"true"`，設定為 `"true"`。
+  - 只有真的變更 ConfigMap 時才重啟 `argocd-redis`、`argocd-server` 與 `argocd-application-controller`。
+- 更新 `k8s/argocd/DEPLOY.md`，新增「套用 Application Health 設定」段落與驗證指令。
+
+**Live 驗證**：
+- `bash scripts/ops/apply-argocd-health-config.sh` 已在 live 執行，結果：
+  - `ok argocd-cm: no global /status ignore`
+  - `ok argocd-cmd-params-cm: controller.resource.health.persist=true`
+  - `no change`
+- 因 live 已是正確狀態，本輪沒有觸發 Argo 重啟。
+
+**進度更新**：
+- ArgoCD health truthfulness：`85% -> 92%`；live 與 repo 手順已對齊，尚待後續把 Argo base install 也納入完整 Kustomize/Helm source。
+- GitOps control-plane hygiene：`55% -> 68%`；已補健康設定 source-of-truth，仍需盤點 Argo 安裝來源、RBAC 與 notification config。
+- 完整 AI 自動化飛輪：維持 `67%`；本輪降低 GitOps 可觀測性退化風險，未新增 auto-repair execution。
+
 ## 2026-06-03｜AwoooP Work Items Owner Review Gate 與 Mobile Shell 可讀性
 
 **背景**：統帥要求 AwoooP / AI 治理不能只在 Telegram 噴告警，前端必須看得出事件跑到哪個流程、誰要接手、AI 做了什麼、哪些步驟被 gate 擋住。本階段聚焦 `/zh-TW/awooop/work-items` 的 KM owner-review 接續處理與手機可讀性：把告警中的 `KM 需要更新` 往 Work Items 的單筆審核、乾跑預覽、Owner 確認、寫回保護與 stale ratio 回測串起來。
diff --git a/k8s/argocd/DEPLOY.md b/k8s/argocd/DEPLOY.md
index 88df2b9b..2a7dfbef 100644
--- a/k8s/argocd/DEPLOY.md
+++ b/k8s/argocd/DEPLOY.md
@@ -16,14 +16,34 @@ kubectl apply -f k8s/argocd/argocd-metrics-nodeport.yaml
 kubectl get svc -n argocd | grep nodeport
 ```
 
-## 2. NodePort 配置
+## 2. 套用 Application Health 設定
+
+ArgoCD 必須保留 child resource health，否則 `Application.status.health` 會變成 top-level `Degraded` 但缺少子資源證據。
+
+```bash
+bash scripts/ops/apply-argocd-health-config.sh
+
+kubectl -n argocd get app awoooi-prod \
+  -o jsonpath='{.status.sync.status}{" "}{.status.health.status}{"\n"}'
+```
+
+必要設定：
+
+| ConfigMap | Key | 目標 |
+|-----------|-----|------|
+| `argocd-cm` | `resource.customizations.ignoreResourceUpdates.all` | 不得存在；禁止全域忽略 `/status` |
+| `argocd-cmd-params-cm` | `controller.resource.health.persist` | 必須為 `"true"` |
+
+若腳本實際修改 ConfigMap，會自動重啟 `argocd-redis`、`argocd-server` 與 `argocd-application-controller`，讓設定與 app tree cache 生效。
+
+## 3. NodePort 配置
 
 | Service | NodePort | 用途 |
 |---------|----------|------|
 | argocd-metrics-nodeport | 30882 | Application Controller Metrics |
 | argocd-server-metrics-nodeport | 30883 | ArgoCD Server Metrics |
 
-## 3. Prometheus 抓取端點
+## 4. Prometheus 抓取端點
 
 ```
 http://192.168.0.121:30883/metrics  # Server Metrics (Pod 在 mon1)
@@ -31,7 +51,7 @@ http://192.168.0.121:30883/metrics  # Server Metrics (Pod 在 mon1)
 
 > ⚠️ 注意: ArgoCD Server Pod 運行在 mon1 (121)，需使用該節點 IP
 
-## 4. 關鍵指標
+## 5. 關鍵指標
 
 | 指標 | 說明 |
 |------|------|
diff --git a/scripts/ops/apply-argocd-health-config.sh b/scripts/ops/apply-argocd-health-config.sh
new file mode 100755
index 00000000..a6a985f7
--- /dev/null
+++ b/scripts/ops/apply-argocd-health-config.sh
@@ -0,0 +1,53 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Idempotent source-of-truth for AWOOOI ArgoCD health persistence.
+# Keeps Application child resource health visible and prevents the old global
+# /status ignore rule from making top-level health look degraded/opaque.
+
+NAMESPACE="${NAMESPACE:-argocd}"
+CM_NAME="${CM_NAME:-argocd-cm}"
+PARAMS_CM_NAME="${PARAMS_CM_NAME:-argocd-cmd-params-cm}"
+
+changed=0
+
+get_jsonpath() {
+  kubectl -n "${NAMESPACE}" get cm "$1" -o "jsonpath=$2" 2>/dev/null || true
+}
+
+ignore_all_value="$(get_jsonpath "${CM_NAME}" "{.data.resource\\.customizations\\.ignoreResourceUpdates\\.all}")"
+if [[ -n "${ignore_all_value}" ]]; then
+  echo "remove ${CM_NAME}.data.resource.customizations.ignoreResourceUpdates.all"
+  kubectl -n "${NAMESPACE}" patch cm "${CM_NAME}" \
+    --type=json \
+    -p='[{"op":"remove","path":"/data/resource.customizations.ignoreResourceUpdates.all"}]'
+  changed=1
+else
+  echo "ok ${CM_NAME}: no global /status ignore"
+fi
+
+persist_value="$(get_jsonpath "${PARAMS_CM_NAME}" "{.data.controller\\.resource\\.health\\.persist}")"
+if [[ "${persist_value}" != "true" ]]; then
+  echo "set ${PARAMS_CM_NAME}.data.controller.resource.health.persist=true"
+  kubectl -n "${NAMESPACE}" patch cm "${PARAMS_CM_NAME}" \
+    --type=merge \
+    -p='{"data":{"controller.resource.health.persist":"true"}}'
+  changed=1
+else
+  echo "ok ${PARAMS_CM_NAME}: controller.resource.health.persist=true"
+fi
+
+if [[ "${changed}" == "1" ]]; then
+  echo "restart ArgoCD control-plane to load health config"
+  kubectl -n "${NAMESPACE}" rollout restart deploy/argocd-redis
+  kubectl -n "${NAMESPACE}" rollout restart deploy/argocd-server
+  kubectl -n "${NAMESPACE}" rollout restart statefulset/argocd-application-controller
+  kubectl -n "${NAMESPACE}" rollout status deploy/argocd-redis --timeout=180s
+  kubectl -n "${NAMESPACE}" rollout status deploy/argocd-server --timeout=180s
+  kubectl -n "${NAMESPACE}" rollout status statefulset/argocd-application-controller --timeout=180s
+else
+  echo "no change"
+fi
+
+kubectl -n "${NAMESPACE}" get cm "${CM_NAME}" "${PARAMS_CM_NAME}" \
+  -o jsonpath='{range .items[*]}{.metadata.name}{": ignore-all="}{.data.resource\.customizations\.ignoreResourceUpdates\.all}{", health-persist="}{.data.controller\.resource\.health\.persist}{"\n"}{end}'