Skip to main content

localization-adaptation-method

Produce terminology-controlled, locale-aware content with independent linguistic quality analysis. Use when adapting a fixed source for a named language-locale pair and preparing release evidence.

来源信息

仓库
yangheng95/opencorvus
最近来源活动
2026年8月10日 14:07
检测到的 SKILL.md 语言
英语
星标
312
分支
40

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
localization-adaptation-method
description
Produce terminology-controlled, locale-aware content with independent linguistic quality analysis. Use when adapting a fixed source for a named language-locale pair and preparing release evidence.
# Localization and adaptation method Freeze the source version and locale brief: source language, target language and locale, audience, channel, purpose, tone, formality, glossary, do-not-translate items, variables, markup, units, legal qualifiers, accessibility requirements, and output format. Never infer a locale from language alone. ## Parallel specialist work Run three independent views from the same immutable source scope: 1. Terminology control extracts concepts, preferred terms, forbidden variants, product names, abbreviations, variables, and ambiguous source strings. Every term records source, context, grammatical notes, and approval status. 2. Locale adaptation produces target content and records choices for dates, numbers, currency, units, names, addresses, pluralization, gender, formality, punctuation, typography, examples, cultural references, legal wording, links, and channel constraints. 3. Linguistic quality assurance checks source-target completeness when a target exists and always prepares a locale-specific quality matrix. Check accuracy, fluency, terminology, consistency, grammar, register, locale conventions, placeholders, markup, links, truncation risk, accessibility text, and unresolved ambiguity. The join resolves conflicts against the source, approved terminology, and locale brief. It records accepted changes, open questions, review coverage, and whether a qualified native reviewer participated. Machine-generated or model-reviewed content is never labeled certified human translation. ## Publication boundary The package prepares a release pack and quality evidence. It does not publish, deploy, overwrite a source-of-truth localization system, or claim legal or native-speaker approval without the corresponding authorized evidence.
在 GitHub 查看