菜单
Convert PDF and office documents (DOCX, PPTX, XLSX, HTML) to Markdown by automatically choosing the best tool among Pandoc, Marker, and MarkItDown based on layout complexity, math/equations, and file format strengths.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Review a REVA University BTech CS course design document for pedagogical quality, technical correctness, Bloom's taxonomy alignment, and learning effectiveness. Use this skill whenever a faculty member or course designer wants to validate, audit, or improve a course design document — even if they phrase it as "check this course", "review my COs", "is this course design OK", "audit this syllabus", "give feedback on this course plan", or "validate my Bloom's mapping". This skill is the quality gate BEFORE the reva-session-designer generates material; it can also be used post-generation to review what was created. Input is a course design document in Markdown format. Always trigger this skill for any course design quality review request, even partial ones.
Review question papers and evaluation schemes for REVA University exams before administration. Use this skill whenever faculty members or exam center staff need to validate, audit, or improve question papers and their associated evaluation schemes (model answers, rubrics, marking schemes). Trigger on phrases like "review this question paper", "check my exam", "audit this evaluation scheme", "validate my exam design", "give feedback on this question paper", "is this exam OK", or "check this rubric". Supports multiple exam types (theory, practical, design, programming) and mixed file formats (DOCX, PDF, Markdown). Input should include the question paper and its evaluation scheme. Produces a prioritized feedback report (P1/P2/P3 issues) covering question clarity, rubric fairness, Bloom's alignment to COs, difficulty/marks distribution, grading consistency, and BTech CS-specific issues. Always trigger for any exam quality review request, even partial ones.
Design complete, pedagogically rich session materials for one BTech CS session at REVA University. Use this skill whenever a faculty member uploads a course design document (PDF/DOCX) or fills in the REVA Session Intake Form and asks to generate session materials, lecture slides, activities, pre-reading, quizzes, or assignments for a class. Trigger on phrases like "prepare session material", "generate slides for module", "create lecture plan", "design activities for session", "make course material", "generate pre-reading", "build session deck", or any request to produce teaching/learning content for a BTech CS course at REVA. Always use this skill — even if the request seems partial (e.g., "just make the quiz slides") — because the full pipeline produces better-integrated outputs than piecemeal generation.
Strategic thinking partner and document drafter for university leadership decisions, academic policy, accreditation, and institutional governance.
Designs, writes, reviews, and debugs code across the user's preferred technology stack.
Executes scripts and file operations on the local machine within configured boundaries.
| name | pdftomd |
| description | Convert PDF and office documents (DOCX, PPTX, XLSX, HTML) to Markdown by automatically choosing the best tool among Pandoc, Marker, and MarkItDown based on layout complexity, math/equations, and file format strengths. |
| version | 1.0.0 |
| tags | ["pdf","markdown","converter","pandoc","marker","markitdown"] |
This skill provides decision logic, installation details, code snippets, and a helper script to convert PDF and various document formats into clean Markdown for LLM parsing, RAG pipelines, or document editing. It leverages the relative strengths of three primary tools: Pandoc, Marker, and MarkItDown.
Choose the best tool for conversion by analyzing the input document structure, layout, and complexity:
| Feature / Format | Pandoc | Marker (marker-pdf) | MarkItDown | Recommended Choice |
|---|---|---|---|---|
| Text-Based (Single Column) PDFs | Fair (requires pdftotext first) | Excellent (high accuracy) | Excellent (very fast & clean) | MarkItDown (lightweight/fast) |
| Complex PDFs (Multi-column/Layouts) | Poor | Excellent (restructures columns) | Fair (may merge columns incorrectly) | Marker |
| Math & Equations (PDF) | Poor | Excellent (converts to LaTeX math) | Poor (extracts text chunks) | Marker |
| Scanned PDFs / OCR | Poor | Excellent (native Tesseract/EasyOCR) | Poor (no native image OCR) | Marker |
Word Documents (.docx) | Excellent (very clean headers/tables) | N/A (PDF only) | Excellent (easy API) | Pandoc (cleaner nested structures) |
Excel Spreadsheets (.xlsx) | Fair (converts to basic tables) | N/A (PDF only) | Excellent (cleans cells to MD tables) | MarkItDown |
PowerPoint Slides (.pptx) | Fair | N/A (PDF only) | Excellent (slides-to-text sections) | MarkItDown |
HTML / E-books (.epub) | Excellent (highly customizable) | N/A | Fair | Pandoc |
[Input File Type?]
/ \
[PDF] [Office / Text Docs]
/ \ / | \
[Math/Scanned] [Standard text] [DOCX] [XLSX] [Other]
/ \ / | \
(Marker) (MarkItDown) (Pandoc) (MarkItDown) (Pandoc)
Pandoc is a compiled binary.
winget install JohnMacFarlane.Pandoc
brew install pandoc
sudo apt-get install pandoc
Marker relies on PyTorch and layout models. It requires Python 3.9+ and PyTorch.
pip install marker-pdf
Note: Depending on system configurations, you may also need to install system dependencies such as Tesseract-OCR and ghostscript for scanned PDF support.
Microsoft's lightweight converter can be installed directly from PyPI.
pip install markitdown
A helper script is provided at skills/pdftomd/scripts/convert.py to automate the selection of the best available tool based on the criteria above:
# Auto-select the best tool and convert:
python skills/pdftomd/scripts/convert.py input.pdf -o output.md
# Guide selection for specialized PDFs (e.g. math paper):
python skills/pdftomd/scripts/convert.py paper.pdf -o output.md --pdf-type math
# Force a specific tool:
python skills/pdftomd/scripts/convert.py document.docx -o output.md --tool pandoc
# Convert Word DOCX to GitHub Flavored Markdown (preserves tables and checklists)
pandoc -f docx -t gfm input.docx -o output.md
# Runs the Marker command line tool on a single PDF
marker_single input.pdf --output_dir output_folder/
from markitdown import MarkItDown
md = MarkItDown()
result = md.convert("input.xlsx") # works on pptx, docx, html, simple pdf, etc.
with open("output.md", "w", encoding="utf-8") as f:
f.write(result.text_content)
docu-\nment) and remove them.