style-obfuscatorlisted

# Style Obfuscator Tier 1 of a two-tier design: **flag, don't rewrite.** This version computes a stylometric fingerprint and tells you what's distinctive about a draft's writing style; it never touches the text itself. ## The actual threat model Publishing under your real name on GitHub or LinkedIn isn't itself the risk — that writing is already attributed to you. The risk is that a **distinctive enough public writing fingerprint becomes the reference corpus** an adversary (or, per `privacy- threat-oracle`'s `autonomous_ai_agent` adversary class, an automated scraper doing this at machine scale) uses to match you to something you *didn't* attribute your name to — a pseudonymous post, an anonymous opinion, anything you'd rather keep separate from your professional profile. Hardening the public side makes it a weaker reference for that kind of matching. The most directly useful workflow follows from that: build a reference corpus from your own known writing, then run a **pseudonymous** draft against it before posting — not to check the public draft alone, but to see whether the anonymous one leans on the same distinctive habits (em dash rate, favorite hedges, a specific function-word profile) that a matcher would key on. ## Why this is deterministic-only, not model-based Unlike `privacy-linter`'s deferred stylometric class (genuinely blocked on a local- model decision that hasn't been made — see its `references/leak-taxonomy.md`), computing a stylometric *fingerprint* doe