补充工具动作终态断言的失败诊断

- tests/mod.rs 新增 assert_observation_status，断言失败时输出 tool/status/summary/detail
- command_runtime.rs 中 6 处 observation.status 断言改用该 helper，替换只输出 left/right 的裸 assert_eq
- 目的：CI 上命令类用例失败时能直接看到真实 observation，而不是仅剩状态值差异

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

@@ -1062,7 +1062,7 @@ async fn command_exec_output_sidecar_failure_runs_once_and_requires_reconciliati
     )
     .await;
 
-    assert_eq!(observation.status, "needs-reconciliation");
+    assert_observation_status(&observation, "needs-reconciliation");
     assert!(observation
         .detail
         .as_deref()
@@ -1529,7 +1529,7 @@ async fn agent_runtime_command_exec_revalidates_revision_after_acquiring_project
     )
     .await;
 
-    assert_eq!(observation.status, "verification-failed");
+    assert_observation_status(&observation, "verification-failed");
     assert!(observation
         .detail
         .as_deref()
@@ -1588,7 +1588,7 @@ async fn agent_runtime_command_exec_requires_reconciliation_after_audit_failure(
     )
     .await;
 
-    assert_eq!(observation.status, "needs-reconciliation");
+    assert_observation_status(&observation, "needs-reconciliation");
     assert!(observation.summary.contains("执行结果不完整"));
     assert!(observation
         .detail
@@ -1897,7 +1897,7 @@ async fn agent_runtime_command_exec_agent_db_audit_failure_keeps_gate_failed() {
     )
     .await;
 
-    assert_eq!(observation.status, "needs-reconciliation");
+    assert_observation_status(&observation, "needs-reconciliation");
     assert!(observation.summary.contains("执行审计无法完整落盘"));
     let gate = read_game_creator_agent_runtime_verification_gate(&root, "code-prototype", run_id)
         .expect("read failed audit gate");
@@ -1981,7 +1981,7 @@ async fn agent_runtime_command_exec_rejects_changed_pending_action_identity_afte
         Some(&pending),
     )
     .await;
-    assert_eq!(wrong_id_observation.status, "verification-failed");
+    assert_observation_status(&wrong_id_observation, "verification-failed");
     assert!(wrong_id_observation
         .detail
         .as_deref()
@@ -1998,7 +1998,7 @@ async fn agent_runtime_command_exec_rejects_changed_pending_action_identity_afte
     )
     .await;
 
-    assert_eq!(observation.status, "verification-failed");
+    assert_observation_status(&observation, "verification-failed");
     assert!(observation
         .detail
         .as_deref()

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

@@ -501,6 +501,15 @@ fn wait_for_agent_runtime_lane_release(
     }
 }
 
+/// 断言工具动作 observation 的终态：失败时打印全部字段，避免 CI 只留下 `left == right`。
+fn assert_observation_status(observation: &AgentRuntimeToolObservation, expected: &str) {
+    assert_eq!(
+        observation.status, expected,
+        "observation status mismatch: tool={} status={} summary={} detail={:?}",
+        observation.tool, observation.status, observation.summary, observation.detail
+    );
+}
+
 fn wait_for_agent_runtime_idle(root: &Path, agent_id: &str) -> AgentRuntimeState {
     wait_for_agent_runtime_lane_release(root, agent_id, "wait_for_agent_runtime_idle")
 }