把客戶抽象的 Brief，轉換成
可判斷、可執行的準則

一、把形容詞轉成可觀察的準則

「年輕」「高級」「有活力」這類字，每個人腦裡的畫面都不一樣。第一步不是想方向，是先把形容詞界定成可以驗證的條件。

You are a creative director interrogating a vague brief.

The client used these words to describe what they want: [貼上原話].
Product / service: [是什麼]. Audience: [是誰].

For each vague word, do three things:
1. List 3 different things the client could plausibly mean
2. For each meaning, name one observable, checkable criterion (something
   I could point at in a layout and say yes or no)
3. Write the single question I should ask the client to find out which
   meaning is the real one

Do not propose creative directions yet. Only clarify the words.

輸出裡面最有用的是第三項。那幾條問題直接拿去問客戶，通常一次會議就能把方向收窄一半。

二、找出 brief 裡沒有寫出來的假設

brief 最危險的不是寫錯，是有些事雙方都以為對方知道。

Here is a client brief: [貼上完整 brief].

Identify the assumptions this brief is making but never states. Cover:
- assumptions about the audience
- assumptions about the brand's current perception
- assumptions about what "success" looks like
- assumptions about production constraints

For each assumption, rate how risky it is if it turns out to be wrong
(high / medium / low) and explain the risk in one sentence.

List only assumptions actually implied by the text. Do not invent
context that is not there.

最後那句很重要。不加的話，模型會替你補一堆聽起來合理但完全虛構的背景，你會拿著幻覺去開會。

三、列出你還欠客戶什麼資料

Based on this brief: [貼上 brief]

List the information I still need before I can start creative work.
Group into:
- blockers (cannot start without this)
- shapers (changes the direction significantly)
- nice to have

For each item, write the exact question I should ask, phrased so a
busy client can answer it in one line.

Keep the blocker list to 5 items maximum. If everything looks like a
blocker, the brief is not the problem, my scope is.

最後一句是刻意加的。不設上限的話，你會收到一張二十條的清單，客戶看完直接不回覆。

四、由 brief 推出三個真正不同的方向

大部分人在這一步就直接叫 AI「給我十個 idea」，結果拿到十個同一個方向的變體。要拿到真正不同的方向，要先指定它們在哪一個維度上不同。

You are developing creative territories, not executions.

Brief: [貼上收窄後的 brief]
Evaluation criteria we agreed: [貼上第一條 prompt 得出的準則]

Give me 3 creative territories that differ on a strategic axis, not on
style. For each one:
- the human tension it plays on
- the claim it makes about the brand
- what it deliberately gives up
- one line describing how the first execution would look

Then state which axis separates the three, in one sentence.

Do not give me visual mood or colour yet.

「what it deliberately gives up」那一項是關鍵。一個方向如果什麼都不放棄，代表它其實沒有立場，做出來就會是那種看完沒有感覺的東西。

五、寫一份評分表，用來篩方向

方向多了之後，最怕變成憑感覺挑。有一份寫出來的準則，你跟客戶的討論會從「我覺得」變成「我們之前同意的第三條」。

Turn these agreed criteria into a scoring rubric I can use in a
client meeting: [貼上準則]

Format as a table with:
- criterion
- what a weak answer looks like
- what a strong answer looks like
- weight (must total 100)

Weights should reflect this project's real priority: [寫下最重要的那一項].
Keep it to 6 criteria maximum so it fits on one page.

六、把整份 brief 收成一句話

最後一步。如果收不成一句，代表前面還有東西沒想清楚。

Write a single creative problem statement for this project, using
everything above.

Format: "How might we [do what] for [whom] so that [what changes],
given [the real constraint]?"

Then write three alternative versions that each emphasise a different
part. Tell me which one you would defend and why, in two sentences.

拿到之後，把那句貼回 brief 的第一行。之後每一次有人問「我們到底在做什麼」，答案就在那裡。

怎樣把六條串起來用

順序是有意思的：一到三是把 brief 弄清楚，四是產生方向，五是篩選，六是收束。中間可以來回，但不要跳過一到三就直接做四，那樣你只是在用 AI 更快地做錯的東西。

實際做的時候，把每一步的輸出貼進下一條 prompt 的括號裡。整組跑完大約四十分鐘，比你改到第五版才發現方向錯要快得多。