メニュー
Convert DOCX / PPTX / XLSX / HTML / EPUB to markdown via Microsoft markitdown. Triggers: markitdown, convert document, extract docx, extract pptx, extract xlsx, .docx file, .pptx file, .xlsx file.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | markitdown |
| description | Convert DOCX / PPTX / XLSX / HTML / EPUB to markdown via Microsoft markitdown. Triggers: markitdown, convert document, extract docx, extract pptx, extract xlsx, .docx file, .pptx file, .xlsx file. |
| effort | fast |
| context | inline |
Optional capability built on Microsoft markitdown. Wraps MarkItDown().convert() so the same flow works whether you call it from a skill, the CLI, or scripts.
Read tool doesn't handle/batch plans)/pdf skill (uses Claude Code's built-in Read which has tuned PDF support and pages parameter)ReadTAUSIK convention #19: zero external deps. markitdown is not in the bootstrap manifest. Install on demand:
pip install 'markitdown[all]' # all converters incl. audio/image OCR
pip install 'markitdown[docx,pptx,xlsx]' # narrower install
tausik doc extract reports a clear error if markitdown isn't available — it does not crash TAUSIK.
.tausik/tausik doc extract path/to/file.docx
.tausik/tausik doc extract slides.pptx --format=pptx
Output: markdown on stdout. Exit code 1 if markitdown is missing or extraction fails.
import doc_extract
md = doc_extract.extract_to_markdown("path/to/file.xlsx")
if md is None:
# markitdown missing, file missing, or extraction failed (diagnostic on stderr)
fall_back_to_something_else()
else:
print(md)
extract_to_markdown() never raises — callers can rely on None for graceful degradation.
| Command | Action |
|---|---|
.tausik/tausik doc extract <file> | Convert to markdown |
.tausik/tausik doc extract <file> --format=X | With format hint (logged, markitdown auto-detects) |
python -c "import doc_extract; print(doc_extract.is_available())" | Check if markitdown is installed |
pip install 'markitdown[all]' pulls in heavy deps (Azure cognitive services, OCR). Use [docx,pptx,xlsx] for office-only.head -n 200 before piping to context.Read for small HTML files.brain_post_webfetch hook may eventually use doc_extract for HTML pages — not implemented yet.Auto-test a skill by generating scenarios and running them. Args: [skill-name]. Use when user says 'skill-test', 'test skill', 'validate skill'.
Session-scoped precision mode — read-before-write, verify-before-claim, never-hallucinate-APIs. Use when stakes are high (security, payment, migration). Triggers: 'zero defect', 'precision mode', 'high stakes', 'be careful'.
Read, analyze, and generate Excel/CSV files. Triggers: excel, spreadsheet, csv, export data, create table, .xlsx, .csv.
Read, extract, and analyze PDF documents. Triggers: pdf, read pdf, extract from pdf, analyze pdf, .pdf file path.
Monitor and analyze Sentry errors via MCP. Triggers: sentry, sentry issues, check errors, error tracking, show crashes.