System Prompts for Proofreading And Book Publication Assistants

April 8, 2025 ยท View on GitHub

alt text

April 2025

A collection of system prompts designed for fact-checking, proofreading, and text preparation, aimed at book publishers seeking to automate their editorial processes.

Overview

This repository contains specialized system prompts optimized for Gemini 2.5, which can handle entire book manuscripts and offers a long output window. These prompts are designed to assist publishers with various aspects of the editorial workflow, from proofreading and fact-checking to URL validation and statistical verification.

Key Features

  • Optimized for Gemini 2.5: All prompts are designed to work with Gemini 2.5's capabilities to process entire book manuscripts
  • Structured Output: Many prompts include suggested JSON structures for consistent, machine-readable results
  • Specialized Tools: Categorized prompts for different editorial needs
  • Ready to Use: Copy and paste these prompts directly into your Gemini interface

Categories

Proofreading

  • Typo detection and correction
  • Inconsistency scanning (character names, dates, etc.)
  • General proofreading and copy editing

Statistics Validation

  • Outdated statistics flagging
  • Statistical claim verification
  • Data isolation and verification
  • Currency and exchange rate identification
  • Outdated exchange rate detection

URL Validation

  • Broken link detection and replacement suggestions

English localization

  • British English deviation detection

Content Generation

  • Blurb and tagline ideation

Usage

  1. Select the appropriate system prompt for your task
  2. Copy the prompt text
  3. Paste into Gemini 2.5's system prompt field
  4. Upload your manuscript or paste the relevant section
  5. For prompts with structured output, refer to the corresponding structure.json file in the structured-output folder

Directory Structure

Prompt NameLinkCategoryDescriptionStructured Output
British English DeviationViewEnglish localizationIdentifies deviations from British English, intended for localizing to other variants of EnglishJSON
Find InconsistenciesViewProofreadingAttempts to identify any inconsistencies in the text (highly useful)-
General ProofreaderViewProofreadingFor general proofreading of text, quite open-ended configuration-
General Proofreader 2ViewProofreadingAlternative version of the general proofreader-
Inconsistency ScannerViewProofreadingAlternative version of Find Inconsistencies-
Typo FinderViewProofreadingSelf-explanatory: identifies typographical errors-
Flag Outdated StatsViewStatistics ValidationIdentifies potentially outdated statistics in the text-
FX IdentificationViewStatistics ValidationIdentifies currency and exchange rates in the textJSON
Outdated FX RatesViewStatistics ValidationIdentifies potentially outdated exchange ratesJSON
Stat CheckerViewStatistics ValidationVerifies statistical claims in the textJSON
Stat IsolatorViewStatistics ValidationIsolates statistical claims for verification-
Broken Link CheckerViewURL ValidationIdentifies broken links and suggests replacementsJSON
Blurb IdeationViewContent GenerationGenerates book blurb ideas-
Tagline IdeationViewContent GenerationGenerates book tagline ideas-

Potential Crews for Multi-Agent Frameworks

These prompts can be assembled into different agent teams in multi-agent frameworks. Here are some suggested crews:

Basic Proofreading Crew

AgentRole
Agent OneTypo Finder
Agent TwoGeneral Proofreader
Agent ThreeInconsistency Finder
Agent FourFX Identification

Statistical Validation Crew

AgentRole
Agent OneStat Isolator
Agent TwoStat Checker
Agent ThreeFlag Outdated Stats
Agent FourOutdated FX Rates

Comprehensive Editing Crew

AgentRole
Agent OneTypo Finder
Agent TwoGeneral Proofreader
Agent ThreeInconsistency Finder
Agent FourStat Checker
Agent FiveBroken Link Checker
Agent SixBritish English Deviation (for localization)

Marketing Preparation Crew

AgentRole
Agent OneBlurb Ideation
Agent TwoTagline Ideation
Agent ThreeBroken Link Checker
Agent FourGeneral Proofreader

Each crew can work in sequence or in parallel depending on the multi-agent framework implementation, with outputs from one agent potentially feeding into the inputs of another.