Translate technical & legal documents while preserving original formatting, terminology consistency, and domain context.
文档
Medical Record Translator
试用Use when users need structure-preserving translation of medical records, lab reports, discharge summaries, prescriptions, pathology/radiology reports, or sim...
它能做什么
Use when users need structure-preserving translation of medical records, lab reports, discharge summaries, prescriptions, pathology/radiology reports, or sim...
技能文档
Medical Record Translator
Produce a structure-preserving translation artifact, not a summary report.
Core Rules
- Classify each PDF as
text PDForscanned PDFbefore translation. - Always derive source document blocks first, then translate block-by-block.
- Maintain strict 1:1 mapping: one source block -> one translated block.
- Preserve structure types:
- tables remain tables
- key-value regions remain key-value regions
- free text remains free text blocks
- Output constrained Markdown artifact (not free-form narrative/report writing).
- Chinese-first reading is required, with original source text immediately available for each block.
- Do not add summary-style sections, terminology tables in the main artifact, or explanatory rewrite sections.
- Do not split/merge source paragraphs or regions for readability rewriting.
Operational Workflow
- Ingest file and classify:
text PDF: digital text layer is extractable.scanned PDF: content is image-based or text layer is unreliable.
- Route by type:
text PDFpath: extract text + structural hints first, then map blocks.scanned PDFpath: use a vision-capable model first, perform region detection, then recover region text and block types.
- Build block model from detected structure.
- Translate with one-to-one block alignment and structure preservation.
- Emit constrained Markdown per output contract.
Scanned PDF Mandatory Rules
- For scanned PDFs, parsing must start with a vision-capable model.
- Perform region detection before translation (table regions, key-value regions, paragraph regions, headers, footers as applicable).
- Translate only after region text recovery and block typing are complete.
- Mark uncertainty at the smallest useful scope:
- uncertain cell ownership -> mark that cell or minimal table span
- uncertain key-value field value -> mark that field only
- uncertain phrase -> mark that phrase only
- Do not promote local uncertainty to document-level warnings unless clinically necessary.
Output And Quality References
Follow these documents directly instead of re-specifying them here:
references/output-contract.mdreferences/block-model.mdreferences/quality-checklist.mdreferences/terminology.md
Export
python3 skills/medical-record-translator-publish/scripts/render_translation.py \
skills/medical-record-translator-publish/examples/sample_translation.md \
--output-dir skills/medical-record-translator-publish/examples/rendered-sample
python3 skills/medical-record-translator-publish/scripts/check_rendered_pdf.py \
skills/medical-record-translator-publish/examples/rendered-sample/sample_translation.pdf
Privacy Disclaimer
⚠️ 隐私提醒 | Privacy Notice
医疗病历包含敏感个人信息。在使用本技能前,请注意:
- 请勿上传包含你不愿分享的个人健康信息的文档
- 翻译内容将发送至AI服务进行处理
- 如有隐私顾虑,建议使用本地部署的AI服务或人工翻译
Medical records contain sensitive personal information. Before using this skill:
- Do not upload documents with health information you don't want to share
- Translation content will be sent to AI services for processing
- If you have privacy concerns, consider locally-deployed AI or human translators
相关技能
Generate English hospital recommendation reports for medical travel to China, hospital matching, and redo orders. Use when needs to turn user intake data int...
Framework for translating Word, PDF, Excel, PowerPoint, HTML, Markdown, TXT documents. Provides document parsing code — the AI agent's own LLM performs the a...
Translate PDFs locally by extracting page text and using the agent's own language capability. Use when Codex needs a repeatable workflow for born-digital or...
Complete Chinese and bilingual academic DOCX translation
Translate user-supplied text PDFs into Simplified Chinese and bilingual PDFs. Reads the PDF and only explicitly selected local LaTeX, sends extracted text to an explicitly approved OpenAI-compatible endpoint, runs local PDF/Python subprocesses, and writes outputs, diagnostics, and the default runtime cache under the chosen directory. Local proxy, arXiv download, Docker compilation, and external cache paths require explicit command-line opt-in.