diff --git a/ms_agent/cron/executor.py b/ms_agent/cron/executor.py
index bbae02ee4..68cbdffab 100644
--- a/ms_agent/cron/executor.py
+++ b/ms_agent/cron/executor.py
@@ -4,6 +4,7 @@
 import asyncio
 import contextvars
 import os
+import tempfile
 import time
 from pathlib import Path
 from typing import Any, Optional
@@ -209,5 +210,11 @@ def _write_output(self, job_id: str, output: str) -> None:
         job_dir = self._output_dir / job_id
         job_dir.mkdir(parents=True, exist_ok=True)
         ts = time.strftime('%Y-%m-%d_%H-%M-%S')
-        out_file = job_dir / f'{ts}.md'
-        out_file.write_text(output, encoding='utf-8')
+        with tempfile.NamedTemporaryFile(
+                mode='w',
+                encoding='utf-8',
+                dir=job_dir,
+                prefix=f'{ts}_{time.time_ns()}_',
+                suffix='.md',
+                delete=False) as out_file:
+            out_file.write(output)
diff --git a/tests/cron/test_run_files.py b/tests/cron/test_run_files.py
new file mode 100644
index 000000000..49780ba79
--- /dev/null
+++ b/tests/cron/test_run_files.py
@@ -0,0 +1,72 @@
+"""Preserve every cron result when runs finish in the same second."""
+import asyncio
+from concurrent.futures import ThreadPoolExecutor
+
+import pytest
+
+from ms_agent.cron.executor import JobExecutor
+from ms_agent.cron.manager import JobManager
+from ms_agent.cron.types import CronJobSpec
+
+
+class EchoEngine:
+
+    async def run(self, prompt):
+        await asyncio.sleep(0)
+        return prompt
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize('concurrent', [False, True])
+async def test_same_second_runs_keep_both_outputs(tmp_path, monkeypatch,
+                                                concurrent):
+    monkeypatch.setattr('ms_agent.cron.executor.time.strftime',
+                        lambda *args: '2026-09-22_12-00-00')
+    executor = JobExecutor(output_dir=tmp_path / 'output')
+    monkeypatch.setattr(executor, '_build_engine', lambda *args: EchoEngine())
+    jobs = [
+        CronJobSpec(id='report', prompt=prompt, concurrency=2)
+        for prompt in ['first result', '第二份结果']
+    ]
+
+    if concurrent:
+        results = await asyncio.gather(
+            *(executor.execute(job, config=None) for job in jobs))
+    else:
+        results = [await executor.execute(job, config=None) for job in jobs]
+
+    assert all(result.success for result in results)
+    files = list((tmp_path / 'output' / 'report').glob('*.md'))
+    assert len(files) == 2
+    assert {path.read_text(encoding='utf-8') for path in files} == {
+        job.prompt for job in jobs
+    }
+    if not concurrent:
+        manager = JobManager(tmp_path)
+        assert manager.get_output('report', 0) == jobs[0].prompt
+        assert manager.get_output('report') == jobs[1].prompt
+
+
+def test_multiple_executors_preserve_existing_outputs(tmp_path, monkeypatch):
+    monkeypatch.setattr('ms_agent.cron.executor.time.strftime',
+                        lambda *args: '2026-09-22_12-00-00')
+    monkeypatch.setattr('ms_agent.cron.executor.time.time_ns',
+                        lambda: 1_790_035_200_000_000_000)
+    job_dir = tmp_path / 'report'
+    job_dir.mkdir()
+    old_output = job_dir / '2026-09-22_12-00-00.md'
+    old_output.write_text('existing result', encoding='utf-8')
+    outputs = [f'result {index}' for index in range(8)]
+
+    def write(output):
+        JobExecutor(output_dir=tmp_path)._write_output('report', output)
+
+    with ThreadPoolExecutor(max_workers=4) as pool:
+        list(pool.map(write, outputs))
+
+    assert old_output.read_text(encoding='utf-8') == 'existing result'
+    files = list(job_dir.glob('*.md'))
+    assert len(files) == 9
+    assert {path.read_text(encoding='utf-8') for path in files} == {
+        'existing result', *outputs
+    }