docs: add Localization section stating the separate-repo policy

[jiji262]

Summary

Adds a short ## Localization section to the README, pointing to the Simplified Chinese adaptation. Four lines, one file, no changes to SKILL.md or the 33 patterns. python3 scripts/validate-package.py passes.

## Localization

The patterns here are written for English. Localized variants are maintained as separate community repositories:

- **Simplified Chinese**: [jiji262/humanizer-chinese](https://github.com/jiji262/humanizer-chinese), 36 patterns adapted for Chinese punctuation, phrasing, and register.

Why

Non-English variants are a recurring question on the tracker (#138 Spanish, #163 French, #194 Traditional Chinese, #203 Simplified Chinese) and the README says nothing about it, so every asker opens a new thread and readers looking for their language have no entry point. The wording follows your answer in #163: localized variants stay as separate community repositories, so each language evolves without adding duplicate runtime authorities here.

Background on the linked project

Full details are in #203. The short version: it is an adaptation rather than a translation, because a literal port of the pattern list actively damages correct Chinese.

• Curly quotes (§19) invert. “” are the standard in Simplified Chinese (GB/T 15834-2011), not a tell. That slot detects full-width/half-width punctuation mixing instead, which is the real machine fingerprint.
• The em dash ban (§14) is softened, not copied. The Chinese dash —— is legitimate punctuation for explanation and topic shifts, so only overuse, paired parenthetical dashes, and mis-typesetting (single-width —, or - / -- standing in for a dash) are flagged.
• Subjectless sentences (§13) are idiomatic Chinese, not hidden-actor passives. Flagging them would gut normal writing, so that slot targets Europeanized 被-constructions instead.
• Hyphenated word pairs (§26) are dropped. Chinese has no hyphenated compounds.

Net: 24 patterns map one-to-one, 8 collapse into 4, 1 is dropped, and 8 Chinese-specific patterns are added, for 36 total. The added patterns are sourced rather than invented, mostly from a 6,586-text parallel corpus study (CCL 2023, Beijing Language and Culture University): AI conjunction density runs about 3x human, modal-particle density about 1/5, and antithesis stacking about 6x.

Two additions may generalize past Chinese, since both came out of running "de-AI-ify this" on Chinese AI text without a skill and reading what went wrong:

1. A register gate. For encyclopedic, technical, and reference text, neutral plain prose is the correct human voice. Without an explicit gate, "humanizing" turns a reference article into a personal blog post.
2. An anti-overcorrection list. The most common failure was not leftover AI tone. It was inventing first-person anecdotes ("back when I was learning to code...") to manufacture human texture, which is fabrication wearing a human mask. Your no-fabrication rule from v2.9.0 is extended to name that case explicitly.

MIT, with attribution to blader/humanizer kept in the LICENSE, README, and SKILL.md.

Disclosure

I maintain the linked project, so this is not a neutral contribution and you should weigh it as such.

I also kept the scope deliberately narrow: no fork directory, no table of other people's repos. #59 proposed a "Community Forks" table and you closed it as not matching first-party scope, and I did not want to re-litigate that. If a single third-party link still reads as out of scope, closing this is completely fine. The information lives in #203 either way, and I would rather the README stay yours than have my link in it.