Slop Guard

Programmatic prose linter that detects formulaic writing patterns by running approximately 80 compiled regex rules against input text. Checks overused vocabulary, stock phrases, structural patterns like bold-header-explanation blocks and long bullet runs, tone markers, rhythm monotony, em dash density, contrast pairs, and repeated n-grams. Scoring uses exponential decay with a concentration multiplier that penalizes clustered violations more heavily than diverse ones, producing a 0-100 score with banding from "clean" to "saturated." Exposes two tools: check_slop for inline text analysis and check_slop_file for file-based analysis. Also ships a CLI (sg) for terminal use and CI gating, plus a fitting tool (sg-fit) to train custom rule configurations from corpus data.