by superwyh
Engineering-focused game localization for coding agents: contextual translation, naturalness review, engine integration, QA, and detailed developer reports.
# Add to your Claude Code skills
git clone https://github.com/superwyh/Game-Localization-EngineeringGuides for using ai agents skills like Game-Localization-Engineering.
See how Game-Localization-Engineering compares with popular alternatives.
Game-Localization-Engineering is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by superwyh. Engineering-focused game localization for coding agents: contextual translation, naturalness review, engine integration, QA, and detailed developer reports. It has 73 GitHub stars.
Game-Localization-Engineering's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/superwyh/Game-Localization-Engineering" and add it to your Claude Code skills directory (see the Installation section above).
Game-Localization-Engineering is primarily written in Python. It is open-source under superwyh on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh Game-Localization-Engineering against similar tools.
No comments yet. Be the first to share your thoughts!
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM lists this project for discovery and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
English · 简体中文 · 繁體中文 · 日本語 · 한국어 · Français · Deutsch · Español (ES) · Português (BR) · All languages
Copy this prompt into your coding agent:
Install the game-localization skill from this repository:
https://github.com/superwyh/Game-Localization-Engineering
Locate game-localization/SKILL.md and preserve the complete folder.
Use the current agent’s supported user-level skill directory.
Check existing local changes before updating.
Verify Python 3.10+ and run scripts/check_package.py in the installed package.
Report the installation path, check results, and invocation method.
Game Localization Engineering is a self-contained skill for translating game projects, preserving resource integrity, reviewing updates, and producing a detailed developer handoff. The agent translates using the game’s context; bundled Python tools manage extraction, review records, structural checks, staging, integration, and recovery.
| Capability | What it handles |
|---|---|
| Context and voice | Terminology by concept; speaker, addressee and story-stage voice; UI, dialogue, subtitles, items, quests, humor and linked puzzle clues. |
| Resource protection | Stable IDs, parameters, markup, protected literals, source revisions and metadata. Direct adapters for flat JSON maps, CSV and TSV. |
| Incremental updates | Retain unchanged translations, mark changed source/context/guidance as stale, keep the original translation basis, and preserve removed entries for recovery. |
| Returned reviews | Export CSV packets; match reordered rows by ID; detect competing edits, changed source fields and concurrent local changes before writing. |
| Guarded integration | Stage outputs, inspect per-ID changes, check destination hashes, retain backups and refuse rollback over later developer edits. |
| Coverage and LQA | Separate logical messages, physical resource slots and build-bound scene evidence; validate layout, switching, input, fonts and accessibility when applicable. |
| Scoped compliance | Off by default. Enable triage or release review only for selected markets, platforms and mechanics, using local versioned evidence. |
| Developer delivery | Full source/target comparison plus decisions, file changes, actual checks, unresolved issues, exclusions and a concrete verification checklist. |
Review awkward phrasing, source-language syntax, repetitive templates, and unsuitable register while preserving meaning, character voice, clues, and resource syntax. Establish the meaning, edit in the target language, read the complete scene, then return to the source to check facts, uncertainty, conditions, and game function. Keep good text and intentional repetition unchanged.
For example, The seal may be weakening. can become 封印可能正在减弱。 in simplified Chinese; 封印已经松动了。 changes a possibility into an established fact. Naturalness does not authorize that change. Dynamic grammar, puzzle layout, and subtitle timing may need engineering work.
Use only the requested language and regional rules. Record significant before/after edits, reasons, preserved constraints, actual checks, and developer questions. The CLI records and reports judgments; it does not automatically certify naturalness. AI detector scores are not acceptance criteria. Keep editorial commentary out of shipped strings and final prose, while retaining useful review evidence.
Review this game's existing translations for naturalness. Preserve meaning, character relationships, mechanics, clues, and variables. Keep a before/after record of substantive edits with reasons. Recheck the final wording against the source, validate the applicable resources, and report unperformed checks and developer decisions.
Naturalness procedure · Decision examples and report record
Select the languages the game actually needs. The workflow does not use a fixed global translation list. An unlisted language can use the baseline pipeline, with suitable project guidance, model capability and language review. A documentation edition is not a certification of translation quality for that language.
Included guidance distinguishes mainland simplified Chinese, Taiwan and Hong Kong usage; English regional spelling; Japanese/Korean address and politeness; German/French agreement; Spain/Latin American Spanish; French for Canada; European/Brazilian Portuguese; Russian/Polish inflection; and Turkish/Azerbaijani casing. UI text, voice-over, subtitles, number/date formatting, narration and release markets are configured separately. Character conversion alone does not complete regional localization.
The agent decides which checks to activate from target language/script tags, actual text, player-input scope and the engine’s rendering/input architecture. It then records required, passed, failed, unrun or inapplicable capabilities. The CLI does not automatically diagnose every writing system or certify an engine’s renderer.
| Trigger | Focused checks |
|---|---|
| Arabic, Persian, Hebrew | Mixed RTL/LTR text, names, digits, punctuation, selection and copy; Arabic joining, Persian ZWNJ and Hebrew marks each need appropriate checks. |
| Urdu | Actual Nastaliq font, sloping baselines, ligatures, narrow controls and multiline clipping. |
| Indic scripts | Consonant clusters, reordered vowels, combining marks and safe typewriter/deletion behavior; test each selected script. |
| Thai, Lao, Khmer | Word segmentation, dictionary data in shipped builds, line wrapping and intact clusters. |
| Myanmar | Unicode versus legacy encoding provenance; preserve originals and review conversion decisions. |
| Tibetan | Stacked letters, tsheg, punctuation, line height and syllable/line boundaries. |
| Traditional Mongolian | mn-Mong vertical flow, columns progressing right, suffix NNBSP, caret and selection; separate from mn-Cyrl. |
| Japanese/Chinese vertical text | Enable only for a vertical design; inspect punctuation, ruby, upright digit groups and mixed Latin text. |
| Korean IME and player input | Composition states, Jamo/syllables, focus, undo and save; for global names/chat also test emoji and combining sequences. |
An English/Chinese menu-only project does not acquire an unrelated worldwide script matrix. Global nicknames or chat can trigger display/input checks without translating every menu. String fallback, glyph fallback, shaping and line breaking are tested separately; reversing text or finding a glyph in a font is not sufficient.
The bundled CLI directly edits flat JSON and CSV/TSV. The following engines have integration procedures and acceptance criteria; native formats require the project’s installed editor/exporter or a verified adapter. No automatic all-engine binary importer is claimed.
| Engine / framework | Integration focus |
|---|---|
| Unity | String/Asset Tables, stable shared IDs, Localization/TMP/UI versions, Addressables lifetimes, asynchronous switching and packaged fonts. |
| Unreal Engine | FText namespace/key/source identity; gather → export → import → compile → cook; validate localized assets and packaged culture data. |
| Godot | Translation resources, CSV/PO, context/plurals, tr/tr_n, TranslationServer notifications, TextServer, themes and export data. |
| Yarn Spinner | Stable line IDs and source locks, stale translation updates, line/audio matching and the project’s actual markup parser. |
| Ink | Preserve story logic, tags, glue and original choice indices; presentation refresh must not advance the story or repeat rewards. |
| Ren’Py | Native translate blocks, interpolation, tags, language switching, rollback, skip, saves and one-time side effects. |
| MonoGame / custom | Inspect resource containers, formatters, font shaping, input and save boundaries; implement only the adapter the project needs. |
| Web / Construct | Preserve object/event IDs; separate display language, number formatting, DOM/Canvas layout, focus, narration and IME. |
Integration follows probe → extract → validate → render separately → native import/build → runtime verification. Keep localization lookup and presentation separate from gameplay state. Record unknown formats instead of guessing. Binary/archive projects additionally need encoding, pointers, alignment, compression, unchanged-member checks and build lineage. Existing project frameworks and developer edits remain authoritative.
Use game-localization on this Unity project. The source is en-US; target only zh-Hans-CN, zh-Hant-TW and ja-JP. Preserve the existing String Tables and line IDs. Follow the supplied character and terminology files. Keep compliance off. Translate, review, integrate within the current pipeline, perform available checks, and deliver the full comparison and an English developer report with unrun checks clearly identified.
Refresh against a new source baseline, inspect stale/orphaned entries, then export one locale for review. Reordered rows are matched by ID. A returned row is not approved until its decision explicitly says accepted; conflicting edits reject the entire import before translations are written. REVIEWER must identify the actual reviewer.