Everything you told your AI agent in a prompt is advisory
中文摘要
AI智能体的提示词仅具建议性；其真正的能力边界取决于底层机器的实际环境，而非用户发送的指令。
English Summary
AI agent prompts are merely advisory; an agent's true limits are defined by the underlying machine and its environment, rather than the instructions provided.
Original Excerpt
An agent’s limits are whatever is actually on the machine, not what you sent. Two commands show it, and a map of where your own controls… Continue reading on Medium »