feat: 修复审计落差 #1/#2/#3/#6 + 管理面板移动端适配

#1 工具布尔标记生效(agent-106, 安全红线): loop toolSchemas/runToolCalls 加 allow-set 安全门 restrictSchemas/applyToolSecurityGate——群聊/陌生人只暴露允许工具(默认关), 被拒not_allowed软提示+AuditEvent, owner全工具不变; 身份靠 ActorEnv(非提示词)
#2 checkpoint 恢复(agent-109): run.go LoadLatest + resumableStopReason门控(只恢复in-flight工具循环,不出答案不续跑), 幂等不重放副作用, 新会话不误恢复; resume_test 4用例
#3 tool_result(agent-108, 协议层): Anthropic 原生tool_result块(tool_use_id/is_error)非user纯文本 + OpenAI tool消息补tool_call_id; Content加ToolCallID/IsError等; (loop接线层2处留待,见备注)
#6 capacity_awaken推送(agent-107): onCapacity从只Debug改推WS hub(BroadcastCapacityAwaken, sessionId定向+payload对齐前端AwakenPayload)
管理面板移动端(agent-110): AdminApp 汉堡菜单+抽屉侧栏+标题truncate防换行, RBAC守卫不受影响
全仓 build/vet/test绿 + 前端build绿; 前端无涉差异已对齐

备注: #3 loop接线层2处(assistant工具轮append历史 + Message.ToolCallID/IsError)留给后续; #4代审/#5子代理触发/#7拆文件后置(主人说后三不急)

internal/runtime/loop/tools.go

@@ -114,6 +114,15 @@ func (l *Loop) runToolCalls(ctx context.Context, messages []model.Message, calls
 		res.terminate = true
 	}
 
+	// Security gate (design/11 §3): a group/stranger session may invoke only
+	// tools whose editable marker allows that environment (default closed). A
+	// disallowed call never executes; it is fed back as a not_allowed soft result
+	// and recorded to the audit sink. It is the final authority over the batch —
+	// a later audit hook may tighten but never re-admit a denied tool.
+	if d != nil {
+		l.applyToolSecurityGate(ctx, d, calls, actions)
+	}
+
 	// BatchAdmit hook (audit / political review): block the whole batch.
 	if l.hooks.BatchAdmit != nil {
 		if ok, reason := l.hooks.BatchAdmit(ctx, calls); !ok {