Show HN: Edit Your Books/Essays/Manuscripts with Codex CLI

2026-02-1917:1710github.com

Edit Your Books/Essays/Manuscripts with CODEX CLI. Contribute to jdcampolargo/book_editor development by creating an account on GitHub.

Spotting grammar errors and catching minor inconsistencies in your writing is tedious and frustrating. Sure, you could ask ChatGPT, but it might unintentionally change your original voice. Friends could help but let's face it, they're busy and likely won't get around to it anytime soon. Doing it yourself? You'll definitely miss something. But imagine a tool that swiftly scans your entire manuscript, precisely flags every grammar slip and oversight without altering your style, and provides clear, structured fixes. That’s exactly what I've built using Codex CLI does.

I've been using Codex CLI while editing my book, The Jailbroken Guide to the University, and it makes editing SO MUCH BETTER!!!

With the tool I've built, you get:

  • A detailed table pinpointing exact grammar issues and clear corrections
  • A prioritized action plan for efficient revision
  • Optional full manuscript edits with tracked changes

This repo is designed for non-technical writers and teams. If you can run one command, you can use it.

Example :

  • This repo includes a partial sample from The Jailbroken Guide to the University (about 20,000 words).
  • outputs/issues_table.md: issue-by-issue edits in table form
  • outputs/next_steps.md: what to fix first
  • outputs/clean_tracked.md (optional): tracked-style full edits

Use one of:

  • inputs/manuscript.txt
  • inputs/manuscript.md
  • inputs/manuscript.pdf
  • inputs/sections/*.txt (multi-file mode)

A sample excerpt is already included, so you can test immediately.

If your file is a PDF, run this first:

  • outputs/issues_table.md
  • outputs/next_steps.md

Optional full tracked pass:

GENERATE_CLEAN_DRAFT=1 ./scripts/run_pipeline.sh

If your local Codex config uses an unsupported reasoning level for your model, force one:

MODEL_REASONING_EFFORT=high ./scripts/run_pipeline.sh

Required:

  • codex CLI available in terminal
  • find, shasum (already present on macOS/Linux)

Optional:

  • pdftotext (for PDF ingestion): brew install poppler
  • pandoc (for PDF report rendering): brew install pandoc

Check setup anytime:

  1. You add a manuscript.
  2. Codex analyzes the content using strict editorial rules.
  3. The pipeline validates the outputs for exact format.
  4. If format is wrong, it auto-runs a repair pass.
  5. You get clean outputs ready for editing work.

Most editing pain is triage: finding all real issues and deciding what to fix first. This harness automates that triage and gives structure.

Estimated impact (80k-word draft, rough range):

  • Manual triage: ~10-26 hours
  • Harness-assisted triage: ~1.5-5 hours
  • Time saved: ~8-21 hours
  • Cost-equivalent saved (at $40-$120/hr): ~$320-$2,520

See full assumptions in:

  • docs/guides/time_and_cost_savings.md

Start here:

  • docs/guides/nontechnical_quickstart.md
  • docs/guides/how_it_works.md
  • docs/guides/marketing_kit.md
  • docs/guides/publish_to_github.md

Run this before any public push:

./scripts/safety_check.sh

Optional (recommended): install local git hooks so safety checks run before each commit.

./scripts/install_hooks.sh
.
├── AGENTS.md
├── README.md
├── docs/
│   ├── editor/
│   └── guides/
├── inputs/
├── outputs/
├── prompts/
└── scripts/
  • It preserves voice by design and avoids rewrite-for-taste behavior.
  • Unverifiable facts should be marked with VERIFY: instead of guessed.

Read the original article

Comments

HackerNews