# The Story Bible Starter
The character entry, the who knows what table, and the scene turn audit.

Two artifacts. The character entry, which is what actually reaches the model, and the who knows what table, which no tool
builds for you and which catches the most common continuity error in AI assisted fiction.

## Character Entry

Keep main entries short. A hundred words is generous. Write them as instructions for performance, not as history. History
goes in the notes with the eye color, in whichever field your tool keeps out of the prompt.

Name

Aliases every name this character
is called, by anyone

Diction register and actual
vocabulary

Rhythm sentence length,
interrupts, finishes, waits

What they will not say and what
they deflect to

Verbal tic, or its deliberate
absence

Always doing while they talk

Notes, kept out of the prompt
appearance, backstory, history

The third axis does more work than the other four combined, and it is the one people leave blank. A person who deflects grief into procedure has
to be put in a room with grief, and the scene writes itself the moment you do it.

## The Who Knows What Table

Characters down the side, secrets and revelations across the top, the chapter each character learns each thing in the cells.
Before you write a scene, read down the column of every secret in play and ask who in this room knows it. Fifteen seconds,
and it catches most of what would otherwise reach a copyeditor.

## Character

"Never" is a real value and you need it, because it catches the character who never finds out and yet behaves as though they did. So is "suspects
from Ch 13, confirms Ch 19." So is "wrongly believes it from Ch 6," which is frequently the most productive cell in the table, because a character
acting confidently on a false belief is most of what plot is made of.

Give the model one row, never the whole table. Thirty words of what this character knows as of this chapter prevents the
single most damaging class of error. The whole table contains everything, and everything in the prompt is everything the
model may use.

## The Scene Turn Audit

A scene turns the value charged condition of a character's life. Something that mattered was one way at the start and is the
other way at the end. If nothing turns, you have written an event, not a scene. Run this over eight to twelve scenes at a time,
never the whole manuscript.

# | VALUEATSTAKE | OPEN | CLOSE | TURNED | WHATTURNEDIT

Watch for two adjacent rows negative on the same value. That is a pacing problem as well as a turn problem. And do not let the model fix what
it finds: asked to make a scene turn, it will invent an argument in which two characters say the quiet part out loud. Diagnose with the machine,
fix by hand.