diff --git a/graphify/__main__.py b/graphify/__main__.py
index 910f5164..c68a1f28 100644
--- a/graphify/__main__.py
+++ b/graphify/__main__.py
@@ -2624,6 +2624,29 @@ def main() -> None:
             cohesion = {}
             gods_data = []
 
+        # Fallback: graph.json carries the per-node community as a node attribute
+        # (`to_json` writes it on every node). The analysis sidecar is the
+        # canonical source — but the post-commit / watch rebuild path doesn't
+        # regenerate it, and `extract` may have its temp files cleaned up. When
+        # that happens, `graphify export html` previously bailed with
+        # "Single community - aggregated view not useful." even though the
+        # per-node attribute had the right data all along. Reconstruct from
+        # the graph itself so downstream subcommands (html, obsidian, wiki,
+        # svg, graphml, neo4j) don't silently produce a degraded artifact.
+        if not communities:
+            reconstructed: dict[int, list[str]] = {}
+            for node_id, data in G.nodes(data=True):
+                cid_raw = data.get("community")
+                if cid_raw is None:
+                    continue
+                try:
+                    cid = int(cid_raw)
+                except (TypeError, ValueError):
+                    continue
+                reconstructed.setdefault(cid, []).append(str(node_id))
+            if reconstructed:
+                communities = reconstructed
+
         labels: dict[int, str] = {}
         if labels_path.exists():
             labels = {int(k): v for k, v in json.loads(labels_path.read_text(encoding="utf-8")).items()}
diff --git a/tests/test_cli_export.py b/tests/test_cli_export.py
index 87b98334..e23c987f 100644
--- a/tests/test_cli_export.py
+++ b/tests/test_cli_export.py
@@ -286,3 +286,86 @@ def test_cluster_only_creates_output_dir_when_missing(tmp_path):
     r = _run(["cluster-only", ".", "--graph", str(graph_src), "--no-viz"], tmp_path)
     assert r.returncode == 0, r.stderr
     assert (tmp_path / "graphify-out" / "GRAPH_REPORT.md").exists()
+
+
+# ── communities-fallback when .graphify_analysis.json is absent ──────────────
+# The watch / post-commit rebuild path only writes graph.json + GRAPH_REPORT.md;
+# it does NOT regenerate .graphify_analysis.json. The full `graphify extract`
+# pipeline also removes its temp files at the end of the run on some skill
+# workflows. In both cases the per-node `community` attribute is intact on
+# every node in graph.json — that's the source of truth `to_json` writes.
+# Without these tests, `graphify export html|obsidian|wiki|svg|graphml|neo4j`
+# silently bails or generates a degraded artifact whenever the sidecar is
+# missing, even though the data is right there.
+
+def test_export_html_falls_back_to_node_community_attribute(tmp_path):
+    """When .graphify_analysis.json is absent, export html should reconstruct
+    communities from the per-node attribute in graph.json rather than bailing
+    out with 'Single community - aggregated view not useful.'.
+    """
+    out = _make_graph(tmp_path)
+    # Simulate the watch-rebuild / cleanup case: graph.json + labels survive,
+    # analysis sidecar is gone.
+    (out / ".graphify_analysis.json").unlink()
+
+    r = _run(["export", "html"], tmp_path)
+    assert r.returncode == 0, r.stderr
+    html = out / "graph.html"
+    assert html.exists(), "graph.html should be generated from the fallback"
+    assert html.stat().st_size > 0
+    # The success message comes from to_html — confirm we're not hitting the
+    # "Single community" bail-out path.
+    assert "Single community" not in r.stdout
+    assert "Single community" not in r.stderr
+
+
+def test_export_html_fallback_recovers_multiple_communities(tmp_path):
+    """Stronger assertion: the reconstructed `communities` dict should have the
+    SAME community count as the analysis sidecar would, so downstream code
+    (aggregation thresholds, member counts) sees identical input.
+    """
+    out = _make_graph(tmp_path)
+
+    # Read the canonical community count from the analysis sidecar
+    analysis = json.loads((out / ".graphify_analysis.json").read_text(encoding="utf-8"))
+    expected_count = len(analysis["communities"])
+
+    # And the count we'd reconstruct from graph.json's node attributes
+    graph = json.loads((out / "graph.json").read_text(encoding="utf-8"))
+    reconstructed_cids = {
+        n["community"] for n in graph.get("nodes", [])
+        if n.get("community") is not None
+    }
+    assert len(reconstructed_cids) == expected_count, (
+        f"reconstruction would lose communities: sidecar={expected_count} vs "
+        f"graph.json={len(reconstructed_cids)}"
+    )
+
+    # Now remove the sidecar and confirm the CLI still succeeds end-to-end.
+    (out / ".graphify_analysis.json").unlink()
+    r = _run(["export", "html"], tmp_path)
+    assert r.returncode == 0, r.stderr
+    assert (out / "graph.html").exists()
+
+
+def test_export_html_no_community_data_at_all_still_succeeds(tmp_path):
+    """If a graph.json was somehow written without any per-node `community`
+    attribute (older versions of to_json, hand-built graphs), the fallback
+    should produce an empty communities dict and the renderer should still
+    not crash. Whether the aggregated view is useful is a separate question.
+    """
+    out = _make_graph(tmp_path)
+    (out / ".graphify_analysis.json").unlink()
+
+    # Strip the community attribute from every node
+    graph_path = out / "graph.json"
+    graph = json.loads(graph_path.read_text(encoding="utf-8"))
+    for n in graph.get("nodes", []):
+        n.pop("community", None)
+    graph_path.write_text(json.dumps(graph), encoding="utf-8")
+
+    r = _run(["export", "html"], tmp_path)
+    # Should NOT crash. It may print a warning and skip rendering, but exit
+    # code stays clean — same behaviour as the pre-fallback empty-communities
+    # path, just no longer silently failing on the common case.
+    assert r.returncode == 0, r.stderr