Multilingual Technical Document Translation Engine

Problem

Technical documents moving across languages often fail in practical ways: terms drift, standards are rewritten, and numeric units become inconsistent. Drawings, FMEA sheets, specifications, and standards clauses need translated field names while dimensions, tolerances, part numbers, and standard codes must remain unchanged. A generic LLM workflow may not reliably preserve glossary consistency and layout-sensitive content.

How It Works

The skill runs a staged workflow:
- Phase 0: confirm source and target language, document type, length, and delivery format, then flag standards, acronyms, quantities, units, and figure/table references.
- Phase 1: extract candidate terms, match them against domain glossaries for mechanical, electrical, quality, materials, software, and automotive fields, and produce a project glossary.
- Phase 2: translate by natural paragraphs while preserving ISO, GB, DIN, formulas, symbols, and numeric units.
- Phase 3/4: verify bilingual alignment, scan terminology consistency, review target-language style, and gate output with completeness, numeric integrity, and glossary-rate checks.
- Phase 5: export plain text, aligned bilingual text, Markdown tables, docx, or glossary files.

Boundaries

It is intended for engineering and technical documents, not literary translation, casual dialogue, real-time interpretation, or unprocessed scanned images. For drawing images, run OCR first, then use this skill on the extracted structured text.