diff --git a/.openclaw-sync/source.json b/.openclaw-sync/source.json
index 4809cf31a..2e30b7710 100644
--- a/.openclaw-sync/source.json
+++ b/.openclaw-sync/source.json
@@ -1,5 +1,5 @@
 {
   "repository": "openclaw/openclaw",
-  "sha": "2cd2732ab6695a4b3220ce41166b74aca7097c5f",
-  "syncedAt": "2026-04-25T05:47:24.343Z"
+  "sha": "cfb551c7091d409f38dd447e9d649acd41803aee",
+  "syncedAt": "2026-04-25T05:50:04.202Z"
 }
diff --git a/docs/reference/prompt-caching.md b/docs/reference/prompt-caching.md
index 794a3b8e9..eb15d11bc 100644
--- a/docs/reference/prompt-caching.md
+++ b/docs/reference/prompt-caching.md
@@ -136,6 +136,11 @@ model refs, `contextPruning.mode: "cache-ttl"` is allowed because OpenRouter
 handles provider-side prompt caching automatically. OpenClaw does not inject
 Anthropic `cache_control` markers into those requests.
 
+DeepSeek cache construction is best-effort and can take a few seconds. An
+immediate follow-up may still show `cached_tokens: 0`; verify with a repeated
+same-prefix request after a short delay and use `usage.prompt_tokens_details.cached_tokens`
+as the cache-hit signal.
+
 If you repoint the model at an arbitrary OpenAI-compatible proxy URL, OpenClaw
 stops injecting those OpenRouter-specific Anthropic cache markers.