slop-lint

Purpose

Deterministically measure how much a piece of prose relies on AI-writing tells — negative parallelism, significance inflation, uniform rhythm, formatting reflexes, era-tagged vocabulary, leaked chatbot artifacts — and report span-anchored findings plus a 0-100 cluster score. The script is pure Python stdlib: no network, no LLM, same input → same output. It fixes the observed failure of prompt-only "humanizer" packs: judgments by feel, no numbers, no CI gate, and single-tell overreactions (an em dash is not a verdict).

When to use

• "score / scan / lint / check this text for AI tells or slop"
• "how AI-sounding is this?" — anything wanting a number or a report
• Gating docs or marketing copy in CI before publishing
• Producing the findings a rewrite pass will edit against