diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index 229e07e..e59a327 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -229,6 +229,11 @@ class AnswerEngine:
                 classify["needs_snapshot"] = True
             if re.search(r"\b(how many|count|number of|list)\b", normalized.lower()):
                 classify["question_type"] = "metric"
+            hottest_terms = ("hottest", "highest", "lowest", "most")
+            metric_terms = ("cpu", "ram", "memory", "net", "network", "io", "disk", "load", "usage")
+            lowered_question = normalized.lower()
+            if any(term in lowered_question for term in hottest_terms) and any(term in lowered_question for term in metric_terms):
+                classify["question_type"] = "metric"
 
             if classify.get("follow_up") and state and state.claims:
                 if observer:
diff --git a/atlasbot/llm/prompts.py b/atlasbot/llm/prompts.py
index c469f39..c65dbf4 100644
--- a/atlasbot/llm/prompts.py
+++ b/atlasbot/llm/prompts.py
@@ -24,6 +24,7 @@ ROUTE_SYSTEM = (
     + " Route the question to the best sources and answer style. "
     + "Assume questions are about the Titan Lab Atlas Kubernetes cluster unless the user explicitly asks about something else. "
     + "Prefer snapshot evidence when available. "
+    + "If the question asks for hottest/highest/lowest CPU/RAM/NET/IO/Disk nodes, mark it as a metric question and require snapshot evidence. "
     + "Return JSON only."
 )