What it does

Smell Check is an installable writing skill for AI agents. Review prose for vague claims, inflated language, and wording that has not earned its place.

An agent reads the Markdown and edits. Nothing scans the tree. v1 has no CLI. The TypeScript export is readRule, a file loader, not a checker.

Get started — install the skill in your agent, then use it.

Written Smell Check. Prose, not code smells.

Before / after

Before. This release is a transformative step that will fundamentally change how operators recover from a failed deploy.

After. This release adds a retry button on the failed-deploy screen. Operators can rerun the last job without opening a terminal.

The after version keeps the same fact. It drops unearned elevation. core.md treats that register as a smell, not as a list of forbidden words.

Kept. The 2025 rewrite was transformative: checkout went from three forms to one.

How it fits

core.md          shared law
genre files      surface extras only
project overlay  pronouns, carve-outs, protected lines
                 ↓
drafting
                 ↓
audit.md         smell
claims.md        substance first, then voice

Genre files assume core.md. Do not fork core.md.

Boundaries

Smell Check does not guess whether a human or a model wrote the sentences, and it does not try to fool detectors. Same standard either way. About.

Host pointer — paste agents.md into AGENTS.md or CLAUDE.md. File map: Files.