voice

Voice

A voice is how the user sounds. A style is the shape a platform expects (see reply-styles). A reply has one style and one or more voices.

A voice skill

Each voice is its own skill, named voice-<name>, with these sections and nothing else:

---
name: voice-<name>
description: <one line: when this voice fits>
---

# Voice: <name>

One voice among several: combine it with others as the `voice` skill says, and draft by the
rules and format in `reply-styles`.

## Tone
<two or three lines: how it should feel to the reader>

## Length
<a target, e.g. "two sentences", "as short as it can be">

## Do
- <habits: open with the answer, name the next step, use first names...>

## Don't
- <habits to avoid: apologizing, hedging, exclamation marks...>

## Words
- Use: <words and phrases the user really says>
- Avoid: <words the user never says>

## Example
<one short reply in this voice>

To create a voice, ask the user for two or three replies they were happy with, and draw the sections from those, not from adjectives. Show the draft skill before saving it.

Composing voices

When a request names several voices, such as direct + warm:

1. Apply them in the order given. Where two voices disagree, the later one wins, for that point only.
2. Length comes from the shortest voice named, unless a later voice says otherwise.
3. Don't lists add up: a habit any named voice avoids is avoided.
4. Words add up the same way. If one voice uses a word that another avoids, avoid it.
5. The style's format always wins over every voice: a voice never turns a Slack reply into a letter.

When no voice is named, use the caller's default voice if they gave one, otherwise voice-direct.

Rules

The drafting rules in reply-styles apply to every reply, in any voice.