What problem does it solve? Machine-style translation often produces literal word-for-word output that breaks placeholders, mixes registers, and reads unnaturally in the target language. This Skill performs context-aware translation that infers register, domain, and style from the source text and surrounding codebase, then verifies output against mechanical checks and a quality rubric. ## Core Features & Use Cases - Context-Aware Translation: Infers register, domain terminology, and style from existing locale files, sibling translations, and surrounding code before translating. - Multi-Stage Verification: Runs mechanical checks (placeholder integrity, structure parity, CJK em-dash restructuring, sibling-pattern matching) plus a 50-point translation rubric and anti-AI-pattern scan before emitting output. - Batch and Diff-Sync Modes: Translates batches of i18n keys with terminology consistency, or patches existing translations against an English source diff while leaving untouched sections byte-identical. - Use Case: You updated your English README and need the Korean and Japanese versions brought back in sync. Provide the source diff and target file paths, and the Skill localizes only the changed sections while matching each file's existing voice and terminology. ## Quick Start Translate this UI string file into Korean, preserving all placeholders and matching the register of the existing locale files.
