korean-middle-korean

korean-middle-korean is a creative writing / text-style skill that deterministically rewrites modern Korean text into a Middle-Korean-flavored style, adding archaic particles and endings, Hanja hints, and tone-mark flavor. It is explicitly a creative style transformation, not scholarly restoration — useful for meme-style, Hunminjeongeum-flavored, or old-Korean-voice rewrites, while protecting structural tokens like URLs, emails, Markdown links, and code spans from transformation. The distributed SKILL.md is a thin CLI stub: the primary instructions are not in the file itself but are fetched at runtime.

The skill's required first step is to run a remote command (npx -y @nomadamas/k-skill@0 instruct korean-middle-korean) and follow its printed output as the primary instructions; the CLI detects the runtime and prints applicable instructions, and a bundled Node script (scripts/korean_middle_korean.js) performs the deterministic rewrite via an exec surface, returning JSON with the output text and a replacements log. The bundled instruction.md documents when to use and not use it (avoid academic, legal, medical, or harassment contexts), prerequisites (Node 18+, no API key), a workflow, CLI examples, and an output schema. The stub also carries hard safety rules: never execute payments, message delivery, submissions, cancellations, or public posting without explicit prior approval; never request, print, or store plaintext credentials; and never bypass legal, CAPTCHA, identity-proofing, or e-signature boundaries.

It targets Korean-speaking writers, meme creators, and hobbyists who want a playful archaic Korean voice. The primary security consideration is not malicious behavior but the design pattern of fetching a remote npm package at runtime and treating its output as the authoritative instructions to follow, which is a supply-chain and dynamic-instruction dependency worth noting.