---
name: evidentiary-edit
description: Use this skill when an editor is considering a consequential edit to a manuscript and wants it tested against evidence before it delays release: a structural change, a claim change, a positioning change, or a change of voice. Triggers on requests like "is this edit worth making," "run the evidentiary editing questions," "should I restructure this," or "does this change justify delay."
---

# Evidentiary Edit
This skill carries out the six evidentiary editing questions and the
editorial intervention test from "Every Consequential Edit Needs
Evidence" in *The Editor’s Job After AI Is Tasteful Judgment* It gathers the
evidence and does the mechanical part of a change. The editor decides.
It ends with a recommendation and its reasons.

## Before running

1. Get the proposed edit, the passage it touches, and the reason the
   editor gave for it.
2. Get any reader evidence that exists: reviews, beta-reader notes,
   support questions, errata, competitive books. Do not invent any.

## Steps

1. Answer the six questions in writing:
   - What problem does the edit solve?
   - Who has the problem, and what evidence says it matters?
   - What reader or commercial outcome could improve?
   - Could a machine safely do the mechanical part?
   - What human judgment does it need?
   - Is the expected benefit worth delaying release?
2. Run the intervention test. Does the edit improve credibility,
   accuracy, evidence, comprehension, structure, differentiation,
   positioning, reader trust, the author's intent, or production
   quality? Or is the only reason preference, taste, habit, elegance,
   or a wish to make the text sound like the editor?
3. For a change that touches the argument or a claim, list every place
   in the manuscript that repeats or depends on it.
4. Show the smallest change that would do the job. Do not apply it.

## Conclude

End with one of these, and give the reasoning.

- Make the edit now. Name the evidence and the release impact.
- Make it in the next edition.
- Ignore it. Name why no reader is affected.
- Human judgment only. Name the decision only the author or editor can
  make.

Where evidence is too thin to recommend, say so and name what would
resolve it. Do not pick an option to sound useful.