fix: address remaining PR review comments — null guard, markup escaping, empty criteria

- Add null check after _load_agent() in benchmark runner (agent can return None on circular refs)
- Escape user-sourced content in Rich markup via _safe_render() in memory panel and skills list
- Default to passed=True when benchmark case has neither expected nor criteria

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lib/cli/src/crewai_cli/agent_tui.py

@@ -1120,7 +1120,7 @@ class AgentTUI(App[None]):
 
         lines = [f"[bold]Active Skills[/] ({len(active)})"]
         for s in active:
-            lines.append(f"  [{_CORAL}]{s.name}[/] — {s.description}")
+            lines.append(f"  [{_CORAL}]{_safe_render(s.name)}[/] — {_safe_render(s.description)}")
         self._mount_sys("\n".join(lines))
 
     def _handle_tasks_command(self, parts: list[str]) -> None:
@@ -1219,13 +1219,13 @@ class AgentTUI(App[None]):
             if mem_type == "canonical"
             else f"[dim]{mem_type}[/]"
         )
-        importance_tag = f" [yellow]★{importance}[/]" if importance else ""
-        scope_tag = f" [{_DIM}]scope:{scope}[/]" if scope else ""
+        importance_tag = f" [yellow]★{_safe_render(str(importance))}[/]" if importance else ""
+        scope_tag = f" [{_DIM}]scope:{_safe_render(str(scope))}[/]" if scope else ""
         time_tag = f" [{_DIM}]{timestamp}[/]" if timestamp else ""
 
         return [
             f"  {i}. {type_tag}{importance_tag}{scope_tag}{time_tag}",
-            f"     {content}",
+            f"     {_safe_render(content)}",
             "",
         ]
 

lib/cli/src/crewai_cli/benchmark.py

@@ -304,6 +304,29 @@ async def _run_model_benchmark(
                     score=0.0,
                 )
 
+            if agent is None:
+                emit(
+                    {
+                        "type": "case_done",
+                        "model": model,
+                        "case_index": i,
+                        "total_cases": total,
+                        "passed": False,
+                        "score": 0.0,
+                        "time_ms": 0,
+                        "error": "agent loader returned None",
+                    }
+                )
+                return BenchmarkResult(
+                    case_index=i,
+                    input=case.input,
+                    expected=case.expected,
+                    actual="[Agent loader returned None]",
+                    model=model,
+                    passed=False,
+                    score=0.0,
+                )
+
             start_ms = _current_time_ms()
             try:
                 response = await asyncio.wait_for(
@@ -365,6 +388,8 @@ async def _run_model_benchmark(
                 )
 
             passed, score = False, 0.0
+            if case.expected is None and case.criteria is None:
+                passed, score = True, 1.0
             if case.expected is not None:
                 passed, score = _check_expected(case.expected, actual)
             if case.criteria is not None: