diff --git a/docs/CLI.md b/docs/CLI.md
index 7f7a685..d50d1f8 100644
--- a/docs/CLI.md
+++ b/docs/CLI.md
@@ -86,6 +86,10 @@ mutiny run --no-hosted   # same as default local
 
 Replay regressions under `.mutiny/tests/` and print a PASS / FAIL / SKIPPED report.
 
+Any failed or skipped case makes the command exit with code `1`, even if other
+cases pass. This includes corrupt regression files and replay errors; an
+incomplete run must not appear successful in CI.
+
 | Arg / flag | Default | Meaning |
 |---|---|---|
 | `regression_id` | (all) | Optional id or name to run one case |
diff --git a/packages/mutiny_cli/src/mutiny_cli/test_cmd.py b/packages/mutiny_cli/src/mutiny_cli/test_cmd.py
index 52ca0d1..272f2bf 100644
--- a/packages/mutiny_cli/src/mutiny_cli/test_cmd.py
+++ b/packages/mutiny_cli/src/mutiny_cli/test_cmd.py
@@ -66,7 +66,7 @@ def run_tests(
     json_out: bool = False,
     write_report: bool = True,
 ) -> int:
-    """Discover and replay project regressions. Exit 0 all pass, 1 failures, 2 error."""
+    """Replay regressions. Exit 0 all pass, 1 failed/skipped cases, 2 setup error."""
     root = project_root.resolve()
     ensure_project_on_path(root)
 
@@ -199,7 +199,7 @@ def run_tests(
 
     if failed:
         return 1
-    if skipped and not passed:
+    if skipped:
         return 1
     return 0
 
diff --git a/tests/unit/test_mutiny_cli_test.py b/tests/unit/test_mutiny_cli_test.py
index 4ffba1d..700470f 100644
--- a/tests/unit/test_mutiny_cli_test.py
+++ b/tests/unit/test_mutiny_cli_test.py
@@ -70,6 +70,35 @@ def test_mutiny_test_fixed_passes(tmp_path: Path):
     assert report["results"][0]["status"] == "PASS"
 
 
+@pytest.mark.parametrize("include_pass", [False, True])
+@pytest.mark.parametrize("contents", [b"{broken", b"{}", b"\xff"])
+def test_mutiny_test_corrupt_artifact_fails(
+    tmp_path: Path, capsys: pytest.CaptureFixture[str], include_pass: bool, contents: bytes
+):
+    _scaffold(tmp_path, fixed=True)
+    cases = tmp_path / ".mutiny" / "tests"
+    if not include_pass:
+        (cases / "refund_limit.json").unlink()
+    (cases / "broken.json").write_bytes(contents)
+
+    assert main(["test", "--path", str(tmp_path)]) == 1
+    assert f"Summary: {int(include_pass)} passed, 0 failed, 1 skipped" in capsys.readouterr().out
+    report = json.loads((tmp_path / ".mutiny" / "test-report.json").read_text())
+    assert (report["passed"], report["failed"], report["skipped"]) == (int(include_pass), 0, 1)
+    broken = next(r for r in report["results"] if r["id"] == "broken")
+    assert broken["status"] == "SKIPPED"
+    assert broken["error"] == "invalid regression JSON: broken.json"
+
+
+def test_mutiny_test_mixed_skipped_json_output(tmp_path: Path, capsys: pytest.CaptureFixture[str]):
+    _scaffold(tmp_path, fixed=True)
+    (tmp_path / ".mutiny" / "tests" / "broken.json").write_text("{broken")
+
+    assert run_tests(project_root=tmp_path, json_out=True) == 1
+    report = json.loads(capsys.readouterr().out)
+    assert (report["passed"], report["failed"], report["skipped"]) == (1, 0, 1)
+
+
 def test_mutiny_test_by_id_and_json(tmp_path: Path, capsys: pytest.CaptureFixture[str]):
     _scaffold(tmp_path, fixed=True)
     code = run_tests(