Slop Audit — Security skill for Claude Code

Slop Audit

Point an AI-slop detector at anything you read. Give it a link, an email, or pasted text, and it tells you how much of it reads like AI slop, with every tell quoted.

Built on top of **Peter Yang's [no-ai-slop](https://github.com/petergyang/no-ai-slop)** skill. Peter's skill removes slop from your own writing and names the patterns it finds. This skill reuses his pattern library and points it outward, at the email, newsletter, article, or post in front of you, and adds ingestion, statistical signals, and a score.

All the pattern-detection credit is Peter's. This skill adds the layer he intentionally left out: reading from a source and scoring it.

What it claims (and what it does not)

It scores **slop density, not authorship.** A human can write slop. A well-edited AI draft can read clean. So the verdict is always "how much this reads like unedited AI," never a hard "a human wrote this" or "a machine wrote this." Every flag quotes the exact line, so you can check it yourself. AI authorship detectors guess; quoted tells are evidence.

Install

Paste this into Claude Code, Codex, or your AI harness:

Install this skill globally: https://github.com/coopersimson96/slop-audit

Then also install Peter's, which this one builds on:

Install this skill globally: https://github.com/petergyang/no-ai-slop

Read from anywhere

• A link: fetched with the firecrawl CLI.
• Your email: read with the Google Workspace CLI (gws) so Claude can scan your inbox.
• Pasted text: drop it in.
• Slack (optional): if you connect a Slack MCP, it audits messages there too. This skill does not set Slack up; it uses it when present.

Use

/slop-audit https://some-newsletter.substack.com/p/whatever

/slop-audit is the last email from that recruiter AI slop?

You get a report card:

SLOP AUDIT  ·  
Score: 82/100 — HIGH (reads like unedited AI)
Confidence: high on slop density, not on authorship
Burs