| name | kordoc-parse-document |
| description | Use when converting Korean document files into readable text with `npx` and kordoc. Covers HWP, HWPX, and PDF parsing to Markdown or JSON, including metadata, outline, warnings, and page count. Trigger for requests such as reading a government document, extracting document contents, summarizing a Korean office file, or parsing a local .hwp/.hwpx/.pdf with `kordoc:parse_document`. |
kordoc: parse_document
Use the public CLI first. This skill maps to the parse_document intent in src/mcp.ts, but without MCP.
Input Requirements
- Require exactly one target document.
- Require the user to attach the document or provide a concrete local path.
- Accept only supported formats:
.hwp, .hwpx, .pdf.
- Do not proceed if the document is missing or the path is ambiguous.
Default Command
npm exec --yes --package=kordoc --package=pdfjs-dist -- \
kordoc /abs/path/document.hwpx
Use JSON when the caller needs blocks or metadata:
npm exec --yes --package=kordoc --package=pdfjs-dist -- \
kordoc /abs/path/document.hwpx --format json --silent
Workflow
- Resolve the target file to an absolute path.
- Confirm the extension is
.hwp, .hwpx, or .pdf.
- Use plain CLI output for human reading.
- Use
--format json when the task needs metadata, outline, warnings, or blocks.
- Summarize the parsed content instead of dumping the whole document unless the user asked for raw output.
Guardrails
- Install
pdfjs-dist alongside kordoc; the package may fail to load without it.
- Preserve warnings about image-based PDFs, skipped elements, or hidden text filtering.
- Do not claim OCR happened unless the execution path explicitly used an OCR-capable custom script.
- If the file is large, summarize key sections first.
- Refuse to continue when no supported document was provided.