Policy — Labeled Multi-Choice Questions

Whenever a single assistant message contains multiple questions *nd*the expected answer is a choice between options, label each option with a letter (abcd) or number (1234) so the user can respond by typing just the label.

Rules

1. Apply both when there's *ne question with multiple options*and when there are *ultiple independent questions*— in the second case, each question becomes a labeled section and each option inside it also gets a label (e.g. 1.a, 1.b, 2.a, 2.b).
2. Do *ot*apply when the expected answer is free text, arbitrary number, file path, or simple yesno confirmation (use `[yN]` in that case).
3. Goal: minimize user typing friction, especially in fast iterative decisions.

Example

❌ Bad:

"Quer que eu use Postgres ou MySQL para o banco? E a autenticação é via Koder ID ou LDAP?"

✅ Good:

*. Banco:*

• 1.a Postgres
• 1.b MySQL

*. Autenticação:*

• 2.a Koder ID
• 2.b LDAP

Responde com, por exemplo, 1.a 2.a.

Why

In iterative technical conversations, typing "Postgres" + "Koder ID" takes longer than typing 1.a 2.a. Multiplied across dozens of decisions per session, the savings are real.