Authored — AI skill for Claude Code

Authored

**Prove you wrote it.** An agent skill for Codex and Claude that helps writers respond when an AI detector flags work they wrote themselves.

It's the opposite of a humanizer: instead of disguising text, it shows the work behind it.


The idea

An AI detector sees	Authored shows your reviewer
One finished document	Every draft, and how each one grew
A probability score	13 recorded days, three drafts, notes and editor metadata
Text that "looks like AI"	The notes and outline each passage came from
Nothing it can verify	Fingerprinted files anyone can check

Detectors misfire on formal, plain and second-language writing, and the companies behind them say their scores shouldn't be the sole basis for a decision. What settles most cases is the writing process itself. Authored finds it, preserves it and builds a readable report.

Try it

**Codex** — install for your user account:

mkdir -p ~/.agents/skills
git clone https://github.com/forrestelliott/authored ~/.agents/skills/authored

In Codex CLI or the IDE extension, select Authored with `/skills` or mention `$authored` in your prompt. If it does not appear, restart Codex. For a project-only installation, use `.agents/skills/authored` inside that project instead. See the [official skill documentation](https://learn.chatgpt.com/docs/build-skills#where-codex-loads-local-skills).

**Claude Code** — install for your user account:

git clone https://github.com/forrestelliott/authored ~/.claude/skills/authored

Then tell your assistant what happened (prefix with `$authored` in Codex to select it explicitly):

Turnitin flagged my history essay as 68% AI and my professor wants to meet Thursday. My drafts and notes are in ~/Documents/history-essay.

The skill guides your assistant to:

1. Preserve yo