diff --git a/evalharness/cli.py b/evalharness/cli.py
index f259547..92a88f1 100644
--- a/evalharness/cli.py
+++ b/evalharness/cli.py
@@ -439,6 +439,14 @@ def _cmd_eval_run(args) -> int:
     from evalharness.data import get_dataset
     from evalharness.viz import render
 
+    comma = [d for d in getattr(args, 'datasets', []) if ',' in d]
+    if comma:
+        tip = ', '.join(comma)
+        fixed = ' '.join(tip.split(','))
+        raise SystemExit(
+            f'error: benchmark names must be SPACE-separated.\n'
+            f'  got:      evalharness eval run {tip}\n'
+            f'  expected: evalharness eval run {fixed}')
     overrides = _overrides(args)
     out_dir = args.out_dir
     if out_dir:
@@ -574,7 +582,7 @@ def _cmd_eval_run(args) -> int:
                 console.print(Panel(
                     f'{type(e).__name__}: {e}',
                     title=f'[bold red]✗ {name} FAILED[/bold red]',
-                    border_style='red', expand=False))
+                    border_style='red', expand=False), justify='center')
             _print_benchmark_result(console, i + 1, total_runs, name,
                                     'failed', _time.time() - t0)