diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4bfb548..35f1138 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - **Contributor issues (wave 3)** — Six new issues from local code study ([#59](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/59)–[#64](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/64)): LENS ghost wikilink nodes, corrupt X-Ray state handling, `agent_write` assert guard, and paired good-first tests.
 - **Contributor issues (wave 4)** — Seven issues from Clean Architecture code study ([#65](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/65)–[#71](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/71)): SYNAPSE cyclic embed duplication, unresolved embed semantics, kinetic dead code, graph watcher DIP, English DX messages, OCP embed refactor, and parametrized SYNAPSE tests. Agent rule: [`.cursor/rules/07-clean-architecture-audit.mdc`](.cursor/rules/07-clean-architecture-audit.mdc).
+- **CLI hygiene** — `scan --broken-refs` now reports unresolved `((uuid))` block references and exits with status 1 when any are found ([#29](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/29)).
 - **Contributor issues (wave 5)** — [#72](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/72) / [#73](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/73): `append_child_to_node` corrupts Markdown when the source file lacks a trailing newline (agent-write data loss).
 
 - **Test coverage (wave 2)** — Community contribution ([#58](https://github.com/MarcoPorcellato/logseq-matryca-parser/pull/58), maintainer #43): **65** new pytest cases (**443** total) for `detect_tab_size_from_markdown`, graph link/backlink helpers, SYNAPSE embedding strip + metadata schema, FORGE Markdown/JSON visitors, LENS node classification, `extract_changelog` CLI, `LogseqConfigReader` timestamps, and KINETIC `agent-write` validation errors. Closes [#20](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/20), [#43](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/43)–[#52](https://github.com/MarcoPorcellato/logseq-matryca-parser/issues/52).
diff --git a/src/logseq_matryca_parser/kinetic.py b/src/logseq_matryca_parser/kinetic.py
index dec46ba..cb2798b 100644
--- a/src/logseq_matryca_parser/kinetic.py
+++ b/src/logseq_matryca_parser/kinetic.py
@@ -149,6 +149,28 @@ def _build_stats_table(pages: list[LogseqPage]) -> Table:
     return table
 
 
+def _build_broken_references_table(
+    graph: LogseqGraph, broken_nodes: list[LogseqNode]
+) -> Table:
+    table = Table(title="Broken Block References")
+    table.add_column("Page", style="cyan")
+    table.add_column("Block UUID", style="magenta")
+    table.add_column("Missing Block Ref", style="bold red")
+
+    for node in broken_nodes:
+        page = graph.page_for_node(node)
+        page_title = page.title if page is not None else "<unknown>"
+        missing_refs = [
+            ref for ref in node.block_refs if graph.get_node_by_embed_ref(ref) is None
+        ]
+        table.add_row(
+            page_title,
+            node.uuid,
+            ", ".join(f"(({ref}))" for ref in missing_refs),
+        )
+    return table
+
+
 def _build_deep_stats_tables(stats: dict[str, Any]) -> tuple[Table, Table, Table]:
     overview_table = Table(title="LENS Deep Statistics")
     overview_table.add_column("Metric", style="cyan")
@@ -222,17 +244,35 @@ def scan(
         None,
         help="Path to the Logseq graph root.",
     ),
+    broken_refs: bool = typer.Option(
+        False,
+        "--broken-refs",
+        help="Print unresolved block references and exit 1 when any are found.",
+    ),
 ) -> None:
     """Scan a graph and print aggregate parsing statistics."""
     resolved = _resolve_graph_path(ctx, graph_path)
 
-    pages = _canonical_pages_from_graph(resolved)
+    from logseq_matryca_parser.graph import LogseqGraph
+
+    graph = LogseqGraph.load_directory(resolved)
+    pages = list(graph.iter_canonical_pages())
     if not pages:
         console.print("[yellow]No Markdown files found under pages/ or journals/.[/]")
         raise typer.Exit(code=0)
 
     console.print(_build_stats_table(pages))
 
+    if broken_refs:
+        broken = graph.get_broken_references()
+        if not broken:
+            console.print("[green]No unresolved block references found.[/]")
+            raise typer.Exit(code=0)
+
+        console.print("")
+        console.print(_build_broken_references_table(graph, broken))
+        raise typer.Exit(code=1)
+
 
 @app.command()
 def visualize(
diff --git a/tests/test_kinetic.py b/tests/test_kinetic.py
index 2dd8a0e..f7d0775 100644
--- a/tests/test_kinetic.py
+++ b/tests/test_kinetic.py
@@ -86,6 +86,23 @@ def test_verbose_flag_enables_debug_logging(tmp_path: Path, caplog: pytest.LogCa
     assert any(record.levelno == logging.DEBUG for record in caplog.records)
 
 
+def test_scan_command_reports_broken_refs(tmp_path: Path) -> None:
+    graph_root = tmp_path / "vault"
+    pages_dir = graph_root / "pages"
+    pages_dir.mkdir(parents=True, exist_ok=True)
+    fake_uuid = "00000000-0000-0000-0000-000000000099"
+    (pages_dir / "Broken.md").write_text(
+        f"- Linker references (({fake_uuid}))\n",
+        encoding="utf-8",
+    )
+
+    result = runner.invoke(app, ["scan", str(graph_root), "--broken-refs"])
+
+    assert result.exit_code == 1
+    assert "Broken Block References" in result.output
+    assert "((00000000-0000-0000-0000-00000" in result.output
+
+
 def test_export_command_json_writes_output_file(tmp_path: Path) -> None:
     graph_root = _create_graph(tmp_path)
     output_dir = tmp_path / "out-json"