fix: normalize runner timeouts

engine/runner.py

@@ -9,22 +9,30 @@ from engine.models import RunResult
 
 def run_command(command: str, cwd: Path, timeout_seconds: int) -> RunResult:
     start = time.perf_counter()
-    completed = subprocess.run(
-        command,
-        cwd=str(cwd),
-        shell=True,
-        capture_output=True,
-        text=True,
-        encoding="utf-8",
-        timeout=timeout_seconds,
-        check=False,
-    )
+    try:
+        completed = subprocess.run(
+            command,
+            cwd=str(cwd),
+            shell=True,
+            capture_output=True,
+            text=True,
+            encoding="utf-8",
+            timeout=timeout_seconds,
+            check=False,
+        )
+        exit_code = completed.returncode
+        stdout = completed.stdout
+        stderr = completed.stderr
+    except subprocess.TimeoutExpired as exc:
+        exit_code = 124
+        stdout = exc.stdout or exc.output or ""
+        stderr = exc.stderr or f"command timed out after {timeout_seconds} seconds"
     runtime_seconds = time.perf_counter() - start
     return RunResult(
         command=command,
         cwd=cwd,
-        exit_code=completed.returncode,
+        exit_code=exit_code,
         runtime_seconds=runtime_seconds,
-        stdout=completed.stdout,
-        stderr=completed.stderr,
+        stdout=stdout,
+        stderr=stderr,
     )

tests/test_execution_pipeline.py

@@ -15,6 +15,16 @@ class ExecutionPipelineTest(unittest.TestCase):
         self.assertEqual(result.exit_code, 0)
         self.assertIn("ok", result.stdout)
 
+    def test_run_command_returns_result_on_timeout(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            result = run_command(
+                "python -c \"import time; time.sleep(2)\"",
+                Path(tmp),
+                timeout_seconds=1,
+            )
+        self.assertNotEqual(result.exit_code, 0)
+        self.assertIn("timed out", result.stderr.lower())
+
     def test_parse_score_output_reads_primary_score(self) -> None:
         score = parse_score_output(
             '{"score": 4.5, "metrics": {"violation_count": 0}}',