Your taste,
not the model’s

AI agents are rats in a maze. They reach for what they know, and what they know is slop.

stupify reviews every PR against a corpus of code you actually respect, names the slop, and tells you how to fix it.

$npx @stupify/cli
View on GitHub ↗

Runs on your own Codex plan. No API keys, no servers you run. MIT licensed.

A stupify review surfacing four kinds of finding in one pass: a bug, a fail-open footgun, a reinvented SDK primitive, and a dead config seam.

stupify holds your standard on both ends of the loop. The same .review/ primes your agent before it writes a line and reviews the pull request after, so the best catch is the slop that never ships.

Every diff is judged against a corpus you choose: a taste pack (code like dtolnay, DHH, or antirez) or your own best files, with a rubric that names what counts as slop and keeps the fix small. It runs on your own Codex plan, not a metered key, so a careful review costs a fraction of the API.

No corpus yet? Borrow one

Pick a programmer whose code you’d point a new hire at, or compose several. Each pack is concrete principles plus commit-pinned exemplar files. Or bring your own best files.

Stop shipping slop

One command. Open Claude Code primed, or open a PR and let it review.

$npx @stupify/cli
Star on GitHub