The /schema output should include the isConsequential flag that OpenAI uses. Since people running this backend on their vault know what they are doing, I suggest setting it to false for all endpoints, removing the necessity to confirm each API call in ChatGPT