diff --git a/backend/agenteval/intelligent_eval/lifecycle.py b/backend/agenteval/intelligent_eval/lifecycle.py
index ee8db62..cf7bca2 100644
--- a/backend/agenteval/intelligent_eval/lifecycle.py
+++ b/backend/agenteval/intelligent_eval/lifecycle.py
@@ -197,9 +197,15 @@ def cancel(session: Session, eval_id: str) -> IntelligentEval:
 def submit_report(session: Session, eval_id: str, report: dict[str, Any]) -> IntelligentEval:
     """OpenClaw 提交结构化报告：executing → completed。
 
-    报告结构校验在路由层（pydantic），此处只负责落库与状态迁移。
+    报告结构校验在路由层（pydantic），此处只负责落库与状态迁移。提交前必须
+    满足：存在会话时须全部 close（completed）——否则拒绝，杜绝"评估 completed
+    但 completed_sessions=0（进度 0%）"的不一致（worker 建会话未 close 时，
+    analyst 不应提前把评估标完成）。
     """
     repo = IntelligentEvalRepository(session)
+    sessions = IntelligentEvalSessionRepository(session).list_by_eval(eval_id)
+    if sessions and any(s.status != IntelligentEvalSessionStatus.COMPLETED for s in sessions):
+        raise IntelligentEvalTransitionError("会话未全部关闭，不能提交报告")
     result = repo._submit_report_if_executing(eval_id, report)
     return _resolve_write(
         eval_id,
diff --git a/tests/integration/test_intelligent_evals_api.py b/tests/integration/test_intelligent_evals_api.py
index a2f36f5..cab77a5 100644
--- a/tests/integration/test_intelligent_evals_api.py
+++ b/tests/integration/test_intelligent_evals_api.py
@@ -537,6 +537,31 @@ class TestReport:
         resp = await client.get(f"/api/intelligent-evals/{eval_id}/report")
         assert resp.status_code == 404
 
+    async def test_submit_report_rejects_when_sessions_not_closed(self, client, seeded_db):
+        """评估不应在会话未全部 close 时被标 completed（否则进度 0% 但状态完成）。
+
+        复现 t480 真实问题：worker 建会话但未 close，analyst 仍提交报告 →
+        submit_report 无条件 executing→completed，completed_sessions=0。
+        """
+        eval_id = await _create_executing_eval(client)
+        await _create_session(client, eval_id)  # 建 1 个 running 会话，不 close
+        resp = await client.put(f"/api/intelligent-evals/{eval_id}/report", json={"report": _report_payload()})
+        # 会话未全部 close → 应拒绝（409），而非 200 直接 completed
+        assert resp.status_code == 409, resp.text
+
+    async def test_submit_report_allows_when_sessions_closed(self, client, seeded_db):
+        """会话全部 close 后提交报告应成功（正常 analyst 路径不被破坏）。"""
+        eval_id = await _create_executing_eval(client)
+        session = await _create_session(client, eval_id)
+        resp = await client.post(
+            f"/api/intelligent-evals/{eval_id}/sessions/{session['id']}/close",
+            json={"verdict": {"goal_achieved": True, "summary": "完成", "severity": "low"}},
+        )
+        assert resp.status_code == 200, resp.text
+        resp = await client.put(f"/api/intelligent-evals/{eval_id}/report", json={"report": _report_payload()})
+        assert resp.status_code == 200, resp.text
+        assert resp.json()["status"] == "completed"
+
     async def test_markdown_before_submission_returns_404(self, client, seeded_db):
         eval_id = await _create_executing_eval(client)
         resp = await client.get(f"/api/intelligent-evals/{eval_id}/report/markdown")