---
name: pacing-engine
description: Audit narrative tempo through scene function, event density, decision frequency, information movement, escalation, consequence visibility, and breathing space, rather than word counts. Use when a draft feels rushed, stalled, or flat.
---
# Pacing Engine v2.1

Evaluate tempo by what each unit does, not how long it is. A slow scene that changes state is working; a fast scene that changes nothing is not.

## What to measure

- **Scene function** — what this scene must accomplish, and whether it does.
- **Event density** — how much actually happens.
- **Decision frequency** — how often characters act rather than react.
- **Information movement** — how much is learned, concealed, or misread.
- **Conflict progression** — whether the situation is genuinely changing.
- **Consequence visibility** — whether the reader can see effects arriving.
- **Scene length relative to function** — whether the unit is the size its job requires.
- **Breathing space** — whether the story allows any recovery.

## Record

Maintain audits in `pacing/PACING.md`, with the function table, findings, and each recommended adjustment's cost elsewhere. An adjustment that removes a payoff discharge is a structural change and needs approval, not a pacing note.

## Detect

- **Stalls** — extended sequences where nothing changes.
- **Repetitive beats** — the same turn, conflict, or beat recurring without variation.
- **Over-compression** — important turns given no room to land.
- **Setup without movement** — a scene that establishes something and ends before anything uses it.
- **Escalation fatigue** — rising stakes that stop registering because nothing is ever lost.
- **Compression after expansion** — the reverse: important material collapsed into summary.

## Two structural traps

- **Front-loading** — spending the arc's tension in the first third, leaving the ending to accelerate. The middle then has nothing to escalate into.
- **Escalation without recovery** — continuous rising pressure with no rest, so the reader adapts and stops feeling it. A scene of genuine safety makes the next danger mean something.

## Method

Recommend structural adjustments rather than arbitrary shortening or lengthening. A scene that is too long is rarely fixed by cutting; it is usually missing a turn, or doing two scenes' work. A scene that is too short usually needs a consequence, not adjectives.

When recommending a change, state which function it serves and what it costs elsewhere — removing a scene frequently removes an obligation discharge.

## Coordination

Use `tension-engine` for pressure within scenes, `plot-engineering` for structural turns, `payoff-debt` to check whether a slow stretch was carrying unresolved promises, and `consequence-engine` when a scene's length is masking missing consequences.