换成好问题

Ask a Better Question is a frequently used prompt stored in unix2dos/brain-prompts. It splits an answer into two turns. First, the model identifies unstated assumptions, missing information that could change the answer, and a common mistake. It then asks exactly one key question and waits. Only after the user replies does it provide the final answer.

Paste the prompt into a new conversation or save it in Alfred, a text expander, or an AI client's prompt library. Use a question that genuinely depends on context for the first check.

Save the prompt named 003_换成好问题 from https://github.com/unix2dos/brain-prompts as a reusable prompt template. Read README.md and lines 19 to 26 of Alfred_Core.md, confirm the name and exact source text, preserve {cursor} as the question placeholder, and name the template Ask a Better Question. Do not rewrite the source prompt. Verify it in a temporary conversation with a non-sensitive comparison question. The first turn must identify assumptions, missing information, and one common mistake, ask one key question, and wait. Do not insert any real private question or sync the template to a cloud service. Ask before overwriting a template, accessing Alfred Snippets, writing global configuration, or enabling cloud sync.

Project URL: https://github.com/unix2dos/brain-prompts/blob/main/Alfred_Core.md#L19-L26
Copy the URL and follow the steps below to complete setup.

1. 1
   Copy the complete prompt

   Keep all three analysis requests, the one-question limit, and the instruction to wait for the user's reply.

2. 2
   Replace the placeholder

   Replace {cursor} with the real question. Do not add unrelated account details, credentials, or confidential information.

3. 3
   Answer the single follow-up

   Review the assumptions and information gaps, then answer only the final question. If the model asks several questions, ask it to reduce them to one.

4. 4
   Continue to the final answer

   Check that the final recommendation uses the new information and states conditions, costs, and uncertainty where relevant.

5. 5
   Optionally save a shortcut

   Store the prompt in Alfred Snippets or a prompt library, preserve {cursor}, and back up any existing template before overwriting it.

This assessment is based on the repository README, the prompt in Alfred_Core.md, and the public file history. The prompt limits clarification to one question while making assumptions and information gaps visible before the answer. That is useful for open-ended requests that otherwise receive generic advice. It also adds at least one turn, the model may choose the wrong key question, and simple requests can become slower. The repository labels it as a frequently used Core prompt but provides no public cross-model evaluation and no license. No model comparison was run for this listing.