Options給我選項

Re-ask whatever is pending as tappable choices, then keep every direction decision clickable for the rest of the session把懸而未決的問題改成可點選的選項重問一次,之後整段對話遇到方向決策都給你點

EN繁中User-invoked使用者叫用
Read SKILL.md on GitHub在 GitHub 看 SKILL.md

Answering an open-ended question by typing is slow, and half the time the answer you'd give is one of three the model already had in mind. This skill turns that around: it re-asks whatever is currently pending as tappable choices, and then keeps every direction decision clickable for the rest of the session.

Install

npx skills add https://github.com/leoluyi/skills -g -a options -y

To update later:

npx skills update options

Source

What it does

Two things, in order.

First it acts on what's already on the table — the last thing asked in prose, or the decision the model was about to make on its own, re-asked right now as selectable choices. Same question, same context, no restarting the topic and no asking you to repeat yourself. If nothing is pending it says so in one line and moves on.

Then, for the rest of the session, any decision that affects direction — architecture, library, data model, scope, sequencing — stops and becomes 2-4 mutually exclusive options instead of a choice made for you. Each is labelled by outcome rather than tone, names its trade-off in one line, and the recommended one comes first and says why. The "Other" escape hatch is added automatically, so it never costs an option slot.

When to use

When you're tired of typing answers to open questions, or when you've noticed decisions being made on your behalf and want them surfaced before they land.

When not to

Not for reversible or mechanical steps — those should just get done, and turning them into questions is its own kind of friction. Not for laying out an entire decision space in full; that's Breakdown. And it steps aside on request: say "autonomous" or "run to completion" and it drops out entirely, batching the decisions into a closing summary instead.

How it works

The rules that make the choices actually usable are about the shape of the option set, not the tone.

Combinations are pre-enumerated. When choices could combine, it proposes "A only", "A + B", "all three" as their own mutually exclusive options, so one click still settles it. Multi-select is reserved for the case where the combinations are too many to list — otherwise the assembly work gets pushed back onto you, which is what you were trying to avoid.

The axis gets covered honestly. One axis per question, both ends listed, including "keep it as is". If a sensible choice is neither listed nor a blend of two that are, the axis was cut wrong and gets redone rather than patched.

Nothing already settled gets re-asked. Anything you've already constrained stays constrained.

Where no tappable-choice tool exists, it falls back to a short numbered list — and plain text is also the right form when the answer is genuinely open-ended: a name, a number, a URL.

Related skills

breakdown is the heavyweight version — it lays out every case at uniform depth before any decision is put to you, where Options just makes the pending one clickable. autopilot is the opposite setting entirely: it explicitly suspends this behaviour and batches every decision into one report at the end.

用打字回答開放式問題很慢,而且有一半的時候,你要給的答案本來就在模型心裡那三個之中。這個技能把它反過來:先把目前懸而未決的問題改成可點選的選項重問一次,之後整段對話遇到方向決策也都給你點。

安裝

npx skills add https://github.com/leoluyi/skills -g -a options -y

之後更新:

npx skills update options

原始碼

它做什麼

兩件事,依序進行。

先處理已經在桌上的——最後一個用散文問的問題,或是模型正打算自己拍板的那個決定,現在就改成可點選的選項重問一次。同一個問題、同樣的脈絡,不重啟話題,也不要你把講過的再講一遍。如果沒有待決事項,它用一行講明然後往下走。

接著,在這段對話的其餘時間裡,任何影響方向的決策——架構、套件、資料模型、範圍、順序——都會先停下來,變成 2 到 4 個互斥的選項,而不是替你選好。每個選項以結果命名而不是以語氣命名,一行講清取捨,建議的那個排第一並說明理由。「其他」逃生口是自動加上去的,所以不會佔掉一個選項的位置。

何時使用

當你不想再打字回答開放式問題的時候,或是你察覺有決策正在替你做、想讓它們在落地前先浮上來的時候。

何時不要

可逆或機械性的步驟不要用——那些直接做就好,硬轉成問題本身就是另一種摩擦。要把整個決策空間完整攤開也不要用,那是 Breakdown。它也懂得退場:你說「自動」或「一路跑完」,它就整個退出,改把決策集中到最後的總結裡。

運作方式

讓這些選項真的好用的規則,重點在選項集合的形狀,不在語氣。

組合預先列好。 當選項可以組合時,它會把「只做 A」「A + B」「三個都做」列成各自獨立且互斥的選項,這樣一次點選還是能定案。多選只保留給組合多到列不完的情況——否則拼裝的工作就被推回給你,而那正是你想避開的事。

誠實覆蓋整條軸。 一題一條軸,兩端都列,包含「維持現狀」。如果有個合理的選擇既不在名單上、也不是名單上兩個的混合,那就是這條軸切錯了,該重切而不是打補丁。

已經定案的不再問。 你已經給過的限制就維持是限制。

在沒有可點選工具的環境,它退回成一份簡短的編號清單——而且當答案本來就是開放式的(一個名稱、一個數字、一個網址),純文字本來就是對的形式。

相關技能

breakdown 是重量級版本——在把任何決策交到你手上之前,它會先把每個案以一致的深度攤開;Options 只是把待決的那一個變成可點選。autopilot 則是完全相反的設定:它會明確暫停這個行為,把每個決策集中成最後一份報告。