Repo: petergyang/no-ai-slop | MIT | 12.4K stars | last push 2026-09-02 | named in a Greg Isenberg video.
What it does
One Markdown skill (a SKILL.md plus an eval file) that tells an AI assistant to edit a draft and remove more than 20 AI-writing patterns while keeping your voice. It can also just detect and flag them without rewriting. No code runs when you use it; a small Python file only zips the package.
Who it is for
Anyone who drafts with an AI assistant and wants a second pass for tells such as “It is not X. It is Y.” reveals, faux-insight setups (“what nobody tells you”) and filler transitions.
Our read
Static review verdict: SAFE. No install hooks, no network calls, no credential reach. Because it is plain text you can read the whole thing in a few minutes.
Honest caveats: if you already use a larger editing skill you will get mostly overlap. We compared it against the one we run and found it covers roughly 80 percent of the same ground; it did add three rules we merged in (colon reveals, a portability test, and deleting a fake-profound last line instead of upgrading it). Pattern checkers also flag plenty of human writing, so treat the output as a prompt to look again, not proof of anything.
How to try it
- Read the repo README and the
SKILL.mdfirst. - Put the skill in your assistant’s skills folder as the README describes.
- Run it on one of your own drafts in detect-only mode and see whether the flags match what you would have cut.
See the other 11 repos
The full vetted list with verdicts, by email.
Sascribe is independent. Figures come from the linked primary pages on the date shown and can change. Nothing here is financial advice or a promise of income.