Slash Command

/init-voice

Build the author's voice/style fingerprint. Hybrid auto-detect — runs the full computational fingerprint over past-books/ if any are present, otherwise runs a 10-question Hebrew interview + sample of the in-progress manuscript. Produces .book-producer/profile.json + AUTHOR_VOICE.md (or AUTHOR_VOICE.draft.md if one already exists).

Invocation

How this command is triggered — by the user, by Claude, or both

Slash command

/hebrew-book-producer:init-voice (no arguments — auto-detects)

Model invocable

No pre-commands

Context Preview

The summary Claude sees in its command listing — used to decide when to auto-load this command

# /init-voice — bootstrap or refresh the author's voice fingerprint

Hand off to the **voice-miner** agent.

## Pre-flight

- `book.yaml` must exist. If not, ask the user to run `/init` first.
- The shared CandleKeep book *Hebrew Linguistic Reference* must be cached at `.ctx/hebrew-linguistic-reference.md`. The `load-candlekeep-guide.sh` script handles this; voice-miner re-runs it if missing.

## Auto-detect

The voice-miner agent counts files in `past-books/`:

- `past-books/*.{pdf,docx,md,txt}` exists with ≥1 file → **heavy path** (full computational fingerprint vs the shared baseline).
-...

Command Content

45 lines · ~507 tokens

Stats

LanguagePython

Parent stars0

MaintenanceGood

Last CommitMay 1, 2026

Actions

View Source View Plugin View on GitHub View README

Stats

Actions

/init-voice — bootstrap or refresh the author's voice fingerprint

Hand off to the voice-miner agent.

Pre-flight

book.yaml must exist. If not, ask the user to run /init first.
The shared CandleKeep book Hebrew Linguistic Reference must be cached at .ctx/hebrew-linguistic-reference.md. The load-candlekeep-guide.sh script handles this; voice-miner re-runs it if missing.

Auto-detect

The voice-miner agent counts files in past-books/:

past-books/*.{pdf,docx,md,txt} exists with ≥1 file → heavy path (full computational fingerprint vs the shared baseline).
otherwise → light path (3-chapter manuscript sample + 10-question Hebrew interview).

What this produces

File	Created by	Editable by hand
`.book-producer/profile.json`	voice-miner	no — re-run `/init-voice` to refresh
`AUTHOR_VOICE.md` (if missing)	voice-miner	yes
`AUTHOR_VOICE.draft.md` (if `AUTHOR_VOICE.md` already exists)	voice-miner	yes — author manually merges into the canonical file

When to re-run

After completing 1–2 chapters of a draft (light → heavy: drop the chapters into past-books/ and re-run).
After significant rewrites that shift voice.
When the linguistic-editor flags a register drift the author wants to relearn.

Report

5-line summary (in Hebrew):

Path used (heavy / light) + what was sampled.
Detected register.
Banned-phrase count + preferred-phrase count.
Burstiness score (and a one-word interpretation: flat / human / very-varied).
Next action — /lector <file> if a draft is ready, otherwise a writing prompt.