TsukiNi22/ai-utils / skills/ux/multilingual-project
multilingual-project
Verifiedi18n - a project in several languages
Request given with the skill: the text after /multilingual-project if any. A website / web app: load multilingual-website too (its web
part: URLs, hreflang, SEO...); this skill gives the common rules.
1. Ask (skill interactive-questions, one or two rounds)
- Languages (multi choice + Other): the ones already there kept; proposals by sets - a few (English + French), the main European ones (en, fr, de, es, it, pt, nl, pl), the most spoken in the world (en, zh-Hans, hi, es, fr, ar, bn, pt-BR, ru, ja, de, ko, tr, vi, id), as many as possible (30+: the sets above + sv, da, nb, fi, cs, el, he, uk, ro, hu, th, ms, fa...). Right-to-left (ar, he, fa, ur) asked explicitly: it changes the layout.
- The source language of the strings (English recommended for the keys / code) and the default for users.
- Who translates: drafted here by the AI and marked "to review" (recommended to start), a translator / a platform (Weblate, Crowdin, Transifex, POEditor: the catalog format follows it), the community.
- Regional variants (fr-FR / fr-CA, pt-PT / pt-BR, en-US / en-GB, zh-Hans / zh-Hant) and what is localised besides the text (dates, numbers, currency, units, sort order, images with text, legal pages).
2. Internationalise the code (once, before any translation)
- Every user-visible string goes to the catalog: UI, messages, errors shown to users, notifications, emails, help, generated files the user reads. Not: logs meant for developers, identifiers, config keys, CLI flags, protocol values, file formats (say it in the recap).
- The standard tool of the stack: Qt
tr()/QT_TR_NOOP+.ts(lupdate / lrelease) · gettext_()+.po(xgettext, msgfmt; C++: libintl or boost.locale) · Pythongettext/ Babel (pybabel extract / update / compile) · Rust Fluent (.ftl) orrust-i18n· Androidres/values-xx/strings.xml· iOS String Catalogs (.xcstrings) · JS / Electron: i18next / ICU MessageFormat JSON · JavaResourceBundle· Gox/text/ go-i18n. - No concatenation of translated pieces: one message with named placeholders (
{count} files in {folder}), the order can change per language. Placeholders documented for translators (comments / context:pgettext,QCoreApplication::translatedisambiguation, ICUdescription). - Plurals with the plural rules of each language (gettext
ngettext+Plural-Forms, ICUplural, Qt%n): Polish / Russian / Arabic have 3-6 forms; gender / select where it changes the words. - Formats from the locale, never by hand: dates and times (and time zones), numbers (decimal / thousands
separators), currencies (symbol position), percentages, units, list joining ("a, b and c"), sorting / collation
(ICU,
std::locale,Intl.Collator,locale.strxfrm), case changes (Turkish dotted i). - Layout: room for +30-50 % longer texts (German, Finnish), no fixed widths around text, wrapping allowed; RTL:
mirrored layout and icons with a direction, logical start / end instead of left / right (Qt
setLayoutDirection, Androidstart/end,supportsRtl), bidirectional text isolated. - Fonts covering every script chosen (Noto families: CJK, Arabic, Devanagari, Thai, Hebrew...), fallbacks set.
- Encoding: UTF-8 everywhere (files, terminals, database, APIs); width of CJK / emoji in terminals (wcwidth).
- Choosing the language: a setting (persisted) > the system / user locale (
LANG,QLocale::system(), Android per-app language,Accept-Languageon a server) > the default; a fallback chain (fr-CA → fr → en) so a missing string never shows a key or a blank; switching without restart when the toolkit allows it. - Pseudo-localisation for tests (accents, +40 % length, brackets:
[Ŝéţţîñĝš ___]) to find hard-coded / clipped text.
3. Translate
- Translations drafted for each language from the source strings, with the context comments, a glossary of the
product terms (kept consistent, product names not translated), the tone of each language (tu / vous, du / Sie),
and each entry marked fuzzy / "needs review" when drafted by AI (gettext
#, fuzzy, Qttype="unfinished", i18next a review flag in the platform). Say clearly which languages need a native review before release. - Never machine-translate legal texts silently: the legal pages in the languages of the market are a
legal-setupmatter (France: French mandatory for consumers - loi Toubon).
4. Check and keep it alive
- A check in the tests / CI (
ci-cd): no missing key per language, no unused key, placeholders identical in every translation, catalogs compiled, no untranslated string left in the code (grep / the extractor's report). - The docs: the README states the languages and how to add one (where the catalogs are, the extract / compile commands, the review flow); a user guide in the languages when asked.
- Recap: languages added (reviewed / drafted), what is not translated and why, how to add a language, the commands.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 193k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 193k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 193k | Updated | Claude CodeCodex | Verified | Install |