稳定AI原生壳终态审计测试

让最终回复失败测试等待 Runtime 终态与 Agent lane 完整释放
让并行后台任务测试在读取对话和审计前等待两个 Agent lane 释放
补充异步终态测试的具体回归用例说明

apps/ai-game-creator-shell/src-tauri/src/tests.rs

@@ -17764,18 +17764,14 @@ async fn background_agent_runtime_marks_response_plan_step_failed_when_final_rep
     )
     .expect("start background task");
 
-    let mut runtime = read_game_creator_agent_runtime_at(&root, "design-director")
-        .expect("read runtime")
-        .state;
-    for _ in 0..750 {
-        if runtime.status == "failed" {
-            break;
-        }
-        std::thread::sleep(Duration::from_millis(20));
-        runtime = read_game_creator_agent_runtime_at(&root, "design-director")
-            .expect("read runtime")
-            .state;
-    }
+    let failed_result = wait_for_agent_runtime_terminal_and_lane_release(
+        &root,
+        "design-director",
+        "design-response-fail-run",
+        "failed",
+        "failed",
+    );
+    let runtime = &failed_result.state;
 
     assert_eq!(runtime.status, "failed");
     assert_eq!(runtime.phase, "failed");
@@ -17787,8 +17783,6 @@ async fn background_agent_runtime_marks_response_plan_step_failed_when_final_rep
         .detail
         .as_deref()
         .is_some_and(|detail| detail.contains("后台 Agent 最终回复调用 LLM 失败")));
-    let failed_result =
-        read_game_creator_agent_runtime_at(&root, "design-director").expect("read failed events");
     let event_types = failed_result
         .recent_events
         .iter()
@@ -46074,24 +46068,22 @@ async fn background_agent_runtime_tasks_can_run_in_parallel_and_persist_replies(
     assert!(combined_requests.contains("后台准备主角规范图"));
     assert!(combined_requests.contains("后台整理玩法循环"));
 
-    let mut art_runtime = read_game_creator_agent_runtime_at(&root, "art-director")
-        .expect("read art runtime")
-        .state;
-    let mut design_runtime = read_game_creator_agent_runtime_at(&root, "design-director")
-        .expect("read design runtime")
-        .state;
-    for _ in 0..250 {
-        if art_runtime.status == "idle" && design_runtime.status == "idle" {
-            break;
-        }
-        std::thread::sleep(Duration::from_millis(20));
-        art_runtime = read_game_creator_agent_runtime_at(&root, "art-director")
-            .expect("read art runtime")
-            .state;
-        design_runtime = read_game_creator_agent_runtime_at(&root, "design-director")
-            .expect("read design runtime")
-            .state;
-    }
+    let art_runtime_result = wait_for_agent_runtime_terminal_and_lane_release(
+        &root,
+        "art-director",
+        "art-background-run",
+        "idle",
+        "completed",
+    );
+    let design_runtime_result = wait_for_agent_runtime_terminal_and_lane_release(
+        &root,
+        "design-director",
+        "design-background-run",
+        "idle",
+        "completed",
+    );
+    let art_runtime = &art_runtime_result.state;
+    let design_runtime = &design_runtime_result.state;
 
     assert_eq!(art_runtime.status, "idle");
     assert_eq!(art_runtime.phase, "completed");
@@ -46105,10 +46097,6 @@ async fn background_agent_runtime_tasks_can_run_in_parallel_and_persist_replies(
         design_runtime.last_response.as_deref(),
         Some("策划后台任务完成：先收敛核心循环。")
     );
-    let art_runtime_result =
-        read_game_creator_agent_runtime_at(&root, "art-director").expect("art runtime result");
-    let design_runtime_result = read_game_creator_agent_runtime_at(&root, "design-director")
-        .expect("design runtime result");
     assert!(art_runtime_result
         .recent_tasks
         .iter()
@@ -46171,9 +46159,6 @@ async fn background_agent_runtime_tasks_can_run_in_parallel_and_persist_replies(
             "assistant conversation must persist before Runtime completes for {agent_id}"
         );
     }
-    drop(wait_to_acquire_agent_runtime_lock(&root, "art-director"));
-    drop(wait_to_acquire_agent_runtime_lock(&root, "design-director"));
-
     fs::remove_dir_all(root).ok();
 }
 

docs/project-memory/shared-memory/pitfalls.md

@@ -35,7 +35,7 @@
 - 现象：isolated child 已显示 idle，单次 all-join reconcile 却偶发返回空列表；或者 Runtime 已显示 completed / failed，Goal、conversation、Agent DB 审计和 per-Agent lock 仍未完成，完整 Rust suite 里出现低概率失败，单独重跑通常通过。
 - 原因：Runtime state、Goal sidecar、终态 result、conversation、审计记录、handoff 清理和执行 lane 释放不是同一个原子观测点；测试只等待 idle / failed 会在同一后台 drain 的 durable 收尾前抢先断言。
 - 处理：产品协议仍以 durable terminal result 和 join readiness 为准。测试在有界时限内等待业务目标终态；需要断言同一 drain 的后续副作用时，同时以 per-Agent runtime task lock 释放为 fence，命中后重新读取投影。join 场景继续重复调用幂等 reconcile，直到取得唯一 join 或超时；不得靠固定长 sleep，也不能因为第一次为空就把协议改成吞掉未完成 child。
-- 验证：`isolated_agents_with_same_template_run_independently_and_join_once` 最多执行 100 次、每次间隔 20ms 的 reconcile，并继续断言只有一个 all-join 和一次父唤醒；Goal、loop-budget、finalization 与 Supervisor reconciliation 测试必须在目标 status / phase 与 Agent lane 同时收束后再读取最终副作用。
+- 验证：`isolated_agents_with_same_template_run_independently_and_join_once` 最多执行 100 次、每次间隔 20ms 的 reconcile，并继续断言只有一个 all-join 和一次父唤醒；Goal、loop-budget、finalization 与 Supervisor reconciliation 测试必须在目标 status / phase 与 Agent lane 同时收束后再读取最终副作用。`background_agent_runtime_marks_response_plan_step_failed_when_final_reply_fails` 和 `background_agent_runtime_tasks_can_run_in_parallel_and_persist_replies` 同样必须经过该 fence 后再断言 `turn.failed` 或 `agent.runtime.completed` 审计。
 - 关联：`apps/ai-game-creator-shell/src-tauri/src/tests.rs`、`apps/ai-game-creator-shell/src-tauri/src/agent.rs`。
 
 ## CI root 环境不能用文件只读权限注入写失败