Localise

English is the canonical source. The target-language deliverable is derived and review-flagged. This skill produces a target-language version of a finished English deliverable, a target-language companion, or a review of your own target-language draft. It does not author from scratch in the target language.

Supported languages: Romanian (ro, fully curated), French (fr), Spanish (es), Russian (ru). French, Spanish, and Russian are working-draft glossaries refined at use. Adding a further language is a new glossary file.

When to use

Trigger for: localising a finished English brief; producing a target-language companion to an English deliverable; reviewing a target-language draft you wrote. Do not trigger for English voice work (use ane-voice) or instrument design (use instrument-designer).

Modes (detected from input)

1. Localise a finished doc then you provide an English deliverable (paste or path) and get a target-language version plus a QA note.
2. English plus companion then Vi or you provide the finished English deliverable and get a target-language companion alongside it.
3. Review my draft then you provide your target-language draft (plus the English source if available) and corrections are walked one by one, never a silent rewrite.
4. Review an image-based deck export then you provide a slide deck or e-learning module exported to Word (Articulate Storyline, PowerPoint), where the translatable text is rasterized inside the slide images and the body holds only slide titles. You get anchored Word comments on a copy plus a branded slide-keyed corrections sheet. Read references/deck-export-mode.md before running it.

Mode 4 triggers when a review target is a .docx whose text lives in images rather than in the body. Confirm it rather than assuming: python scripts/extract_deck.py extract DECK.docx --out-dir slides/ prints the verdict with the evidence behind it, and routes you back to mode 3 when the text is live. The linguistic core (glossary, controlled terminology, back-translation, safeguarding) is identical to mode 3; only the input and output machinery differ.