config.toml配置为： 
approval_policy = “never” 
disable_response_storage = true 
model = “gpt-5.2” 
model_reasoning_effort = “xhigh” 
network_access = “enabled” 
sandbox_mode = “danger-full-access” 
windows_wsl_setup_acknowledged = true 
发比如你好，会有2次请求，cli显示是有回复的，但是看管理面板的日志，总是会有1条错误： 
=== API RESPONSE === 
{“error”:{“message”:“Post "https://chatgpt.com/backend-api/codex/responses\”: read tcp 192.168.64....