Recently, I have added a very useful operation and response rule to the engineering skill set Waza that I use every day, which comes from ASD-STE00, a simplified technical English standard originally used for aviation maintenance documents. At first, I saw Andrej Karpathy's recommendation for the big model to refer to this set of specifications to explain things, and the response would be more readable. Fortunately, I had time tonight and followed through the specifications and the official AI white paper, and found that its requirements for words, conditions, and operation sequence were really suitable for checking whether AI had explained things clearly. Based on my understanding, I wanted to give everyone a better explanation. It can be understood that the aircraft maintenance manual is actually intended for readers from various countries, and English may not necessarily be the other party's native language. For example, when repairing an aircraft, which parts to disassemble, when to operate, and what level to go from the previous step before continuing to the next step, does it feel very similar to now when we ask the agent to do something? We also want the agent to do it as accurately and quickly as possible, and understand our demands in place. It happens to be very similar to the operational guidelines in this aircraft maintenance manual. Many times, changing the meaning of a word or missing a condition in a sentence may result in different actions being taken by the person operating the instruction manual. For example, in a large company, the CEO's ideas go through multiple rounds of transmission, which can easily lead to a lot of information distortion, and the final product may be completely different from what they thought at the time. From my understanding, this set of standards is designed to address the issue of speech and execution bias. It consists of collaborative rules and a dictionary that restricts the parts of speech and meanings that a word can use, including which technologies in the industry are allowed to be used. The same thing will not be renamed to avoid repetition. The steps of the operation need to be clearly written with requirements such as conditions and order, which are actually very easy to understand in technical documents. I have included examples in Waza to help AI better understand, such as in the write writing skill section, to make the content written by AI clearer and more understandable. For example, how to simplify "delete local files after uploading" by AI to "start uploading and then delete local files" can easily lead to problems. The file may be deleted before it is fully uploaded. Here, "complete" determines when it can be deleted and is a necessary word. For example, the phrases "restart only after passing the inspection" and "restart after passing the inspection" meet this requirement, while the phrase "allow only" in this case is the key word that maintains its original meaning. This is exactly what I hope AI can pay attention to when replying and helping me with tasks in daily life. It should explain the results clearly and clearly, so that neither it nor I are confused or confused. When it comes to the point, I don't need to ask any more questions. For example, if the code has been modified, the test has passed, it has been pushed, and the user can download it, it is a detailed description of the results. If it only responds with "fixed and verified", in this case, I will definitely continue to ask whether everything has been resolved. However, I believe that it is not necessary for AI to fully refer to all standards here. The rules summarized are actually very short. "Keep the same name for the same object, specify the object when the reference is unclear, place conditions and exceptions near the corresponding actions, preserve the order relationship, and distinguish between suggestions, requirements, and possibilities. If it is already clear, do not change it. Do not delete reasons and boundaries for the sake of brevity Of course, you can also directly install Waza's Clarity rules to use, or go here to take a look. https://(github.com)/tw93/Waza/blob/main/rules/clarity.md