diff --git a/graphify/__main__.py b/graphify/__main__.py
index 3ff348c..01b69b7 100644
--- a/graphify/__main__.py
+++ b/graphify/__main__.py
@@ -3559,17 +3559,20 @@ def main() -> None:
         else:
             # No labels file yet (or `graphify label` forced a refresh). When run
             # standalone there is no orchestrating agent to do skill.md Step 5, so
-            # auto-name communities with the configured backend rather than leave
-            # "Community N" (#1097). Degrades to placeholders if no backend/on error.
+            # auto-name communities rather than leave "Community N" (#1097).
+            from graphify.cluster import label_communities_by_hub
             from graphify.llm import generate_community_labels
             print("Labeling communities...")
-            # The final labels (LLM or placeholder fallback) are persisted to
-            # .graphify_labels.json by the unconditional write below.
+            # Deterministic, LLM-free base labels: name each community after its
+            # highest-degree hub, so the report is readable even with no backend
+            # (previously bare "Community N"). A configured LLM backend overrides these
+            # with richer names below; its no-backend placeholder fallback does NOT.
+            hub_labels = label_communities_by_hub(G, communities)
             label_communities_input = communities
-            labels = {}
+            labels = dict(hub_labels)
             if missing_only:
                 labels = {
-                    cid: existing_labels.get(cid, f"Community {cid}")
+                    cid: existing_labels.get(cid, hub_labels[cid])
                     for cid in communities
                 }
                 label_communities_input = {
@@ -3581,7 +3584,13 @@ def main() -> None:
                 G, label_communities_input, backend=label_backend, model=label_model, gods=gods,
                 max_concurrency=label_max_concurrency, batch_size=label_batch_size,
             )
-            labels.update(generated_labels)
+            # Only let the LLM OVERRIDE where it produced a real name — its no-backend
+            # fallback returns "Community {cid}" placeholders, which must not clobber
+            # the deterministic hub labels.
+            labels.update({
+                cid: v for cid, v in generated_labels.items()
+                if v and v != f"Community {cid}"
+            })
         stages.mark("label")
         questions = suggest_questions(G, communities, labels)
         tokens = {"input": 0, "output": 0}
diff --git a/graphify/cluster.py b/graphify/cluster.py
index 82f063d..d35c90d 100644
--- a/graphify/cluster.py
+++ b/graphify/cluster.py
@@ -83,6 +83,33 @@ _COHESION_SPLIT_THRESHOLD = 0.05 # re-split communities with cohesion below this
 _COHESION_SPLIT_MIN_SIZE = 50    # only cohesion-split if community has at least this many nodes
 
 
+def label_communities_by_hub(
+    G: nx.Graph, communities: dict[int, list[str]]
+) -> dict[int, str]:
+    """Deterministic, LLM-free community labels: name each community after its
+    highest-degree member — the structural hub — so a report reads ``auth`` /
+    ``log_action`` instead of ``Community 70``. Degree is measured on the full graph
+    ``G``; ties break by node id for run-to-run stability. A community whose members
+    are all absent from ``G`` falls back to ``Community {cid}``.
+
+    Used as the default (no-backend) labeler; an LLM naming pass, when configured,
+    overrides these with richer names.
+    """
+    labels: dict[int, str] = {}
+    for cid, members in communities.items():
+        present = [n for n in members if n in G]
+        if not present:
+            labels[cid] = f"Community {cid}"
+            continue
+        # highest degree wins; ties broken by node id (ascending) for determinism
+        hub = min(present, key=lambda n: (-G.degree(n), str(n)))
+        name = str(G.nodes[hub].get("label") or hub).strip()
+        if name.endswith("()"):
+            name = name[:-2]
+        labels[cid] = name or f"Community {cid}"
+    return labels
+
+
 def cluster(
     G: nx.Graph,
     resolution: float = 1.0,
diff --git a/graphify/watch.py b/graphify/watch.py
index c877036..c3fa5ed 100644
--- a/graphify/watch.py
+++ b/graphify/watch.py
@@ -803,9 +803,12 @@ def _rebuild_code(
         except Exception:
             raw = {}
             labels = {}
-        for cid in communities:
-            if cid not in labels:
-                labels[cid] = "Community " + str(cid)
+        missing = {cid: members for cid, members in communities.items() if cid not in labels}
+        if missing:
+            # Deterministic hub name (highest-degree member) beats a bare "Community N"
+            # placeholder for any community without a saved label.
+            from graphify.cluster import label_communities_by_hub
+            labels.update(label_communities_by_hub(G, missing))
         questions = suggest_questions(G, communities, labels)
         from graphify.report import load_learning_for_report as _llfr
         report = generate(G, communities, cohesion, labels, gods, surprises, detection,
diff --git a/tests/test_community_hub_labels.py b/tests/test_community_hub_labels.py
new file mode 100644
index 0000000..1e50abc
--- /dev/null
+++ b/tests/test_community_hub_labels.py
@@ -0,0 +1,66 @@
+"""Deterministic, LLM-free community labels — `label_communities_by_hub`.
+
+Names each community after its highest-degree member so a report reads "log_action"
+instead of "Community 70", with no backend. Ties break by node id for run-to-run
+stability; a community with no members present in the graph falls back to "Community N".
+"""
+import networkx as nx
+
+from graphify.cluster import label_communities_by_hub
+
+
+def _g(node_labels, edges):
+    g = nx.Graph()
+    for nid, label in node_labels.items():
+        if label is None:
+            g.add_node(nid)
+        else:
+            g.add_node(nid, label=label)
+    g.add_edges_from(edges)
+    return g
+
+
+def test_labels_by_highest_degree_hub():
+    # 'a' is the hub (degree 3); the community is named after it, "()" stripped.
+    g = _g(
+        {"a": "log_action()", "b": "b()", "c": "c()", "d": "d()"},
+        [("a", "b"), ("a", "c"), ("a", "d")],
+    )
+    labels = label_communities_by_hub(g, {0: ["a", "b", "c", "d"]})
+    assert labels[0] == "log_action"
+
+
+def test_not_a_placeholder_for_a_real_community():
+    g = _g({"a": "handler()", "b": "b()"}, [("a", "b")])
+    labels = label_communities_by_hub(g, {0: ["a", "b"]})
+    assert labels[0] == "handler" and labels[0] != "Community 0"
+
+
+def test_tie_breaks_deterministically_by_node_id():
+    # both nodes degree 1 → the lexicographically smaller id wins, regardless of order
+    g = _g({"z": "z()", "a": "a()"}, [("z", "a")])
+    assert label_communities_by_hub(g, {0: ["z", "a"]})[0] == "a"
+    assert label_communities_by_hub(g, {0: ["a", "z"]})[0] == "a"
+
+
+def test_absent_members_fall_back_to_placeholder():
+    # no member of community 5 is in the graph → keep the "Community N" placeholder
+    g = _g({"a": "a()"}, [])
+    assert label_communities_by_hub(g, {5: ["ghost1", "ghost2"]})[5] == "Community 5"
+
+
+def test_node_without_label_attr_uses_id():
+    g = nx.Graph()
+    g.add_nodes_from(["hub", "x", "y"])
+    g.add_edges_from([("hub", "x"), ("hub", "y")])  # hub degree 2, no label attrs
+    assert label_communities_by_hub(g, {0: ["hub", "x", "y"]})[0] == "hub"
+
+
+def test_multiple_communities_each_get_their_own_hub():
+    g = _g(
+        {"h1": "auth()", "a1": "a1()", "a2": "a2()",
+         "h2": "billing()", "b1": "b1()", "b2": "b2()"},
+        [("h1", "a1"), ("h1", "a2"), ("h2", "b1"), ("h2", "b2")],
+    )
+    labels = label_communities_by_hub(g, {0: ["h1", "a1", "a2"], 1: ["h2", "b1", "b2"]})
+    assert labels[0] == "auth" and labels[1] == "billing"