A dependency-free CLI that scans prose for the specific fingerprints of LLM writing — cadence tics, puffery, vague attribution — and reports them as machine-readable linting notes. Agent-first: JSON output, stable schema, and an `explain` command.