- Home
- Skills
- Technical Documentation
- PDF to Markdown: Clean, Faithful Docs for LLMs & RAG
Works with the AI Agents you already use
PDF to Markdown: Clean, Faithful Docs for LLMs & RAG
Convert PDFs, Word, and Excel files to clean Markdown with layout preservation and OCR for RAG and documentation.
Free
PDF to Markdown: Clean, Faithful Docs for LLMs & RAG
Example session with this skill installed
Convert technical_specs.pdf to markdown. It has a two-column layout and several tables. Ensure the tables stay accurate for a RAG system.
- Read your context and instructions
- Compiled the pdf to markdown
I have processed technical_specs.pdf using the marker-pdf engine to handle the two-column layout. The output includes verified tables and LaTeX formulas.
Conversion Report
- Pages: 5
- Converter: marker-pdf
- Tables Found: 2 (Verified for numerical accuracy)
- Assets: 3 images extracted to
assets/ - Integrity Check: Word count variance < 2%.
technical_specs.md
# System Architecture v2.1
<!-- page 1 -->
The following specifications outline the hardware requirements and interface protocols for the primary controller.
## 1. Hardware Overview
The controller operates on a 24V DC bus. All inputs are opto-isolated to prevent ground loops.
| Component | Specification | Tolerance |
|:---|:---|:---|
| Logic Voltage | 3.3V / 5V | +/- 0.05V |
| Operating Temp | -40C to +85C | N/A |
| Flash Memory | 16MB | Fixed |
## 2. Power Calculations
The power draw is calculated using the following formula:
$P = V_{in} \times I_{total} + Q_{loss}$
<!-- page 2 -->
## 3. Pin Mapping
- **J1-1:** Power In (24V)
- **J1-2:** Ground
- **J2-1:** RX Data (UART)
- **J2-2:** TX Data (UART)
[^1]: Ensure all shielding is grounded at the source end only.
Next steps
- Review the extracted images in the
/assetsfolder. - Run the RAG chunking command if you need this prepared for vector ingestion.
- Verify the LaTeX rendering for the formula in Section 2.
Connects securely to your tools. The creator never sees your data.
About this skill
Converting messy office documents into clean Markdown is a frequent bottleneck for RAG pipelines and technical documentation. This skill automates the extraction of text, tables, and images from PDFs, Word, PowerPoint, and Excel files while preserving the original structural hierarchy and formatting.
What it does
- Multi-format extraction converts PDF, DOCX, PPTX, and XLSX files into structured Markdown.
- Layout preservation rebuilds heading hierarchies and handles multi-column text flow.
- Table reconstruction transforms complex document grids into GitHub-flavored Markdown or HTML tables.
- OCR processing handles scanned documents and images to extract embedded text.
- RAG preparation optionally chunks output by headings with metadata front matter for vector databases.
- Formula conversion detects and converts mathematical notations into LaTeX.
How it works
- Analyze source to detect text layers, language, page count, and layout complexity.
- Select engine based on the file type, favoring local tools like
markitdown,marker-pdf, ordocling. - Clean and structure text by removing headers/footers, joining hyphenated breaks, and referencing assets.
- Verify integrity by comparing word counts and spot-checking table data accuracy.
Frameworks & tools
Utilizes specialized CLI tools including Poppler, MarkItDown, Marker-PDF, Docling, and OCRmyPDF.
Why this beats prompting it yourself
Manual extraction via copy-paste loses tables, footnotes, and reading order in multi-column layouts. This skill uses specific parsing engines that understand document geometry, ensuring formulas and metadata remain intact for LLM consumption.
Use cases
- Preparing legacy PDF archives for RAG-based AI search engines.
- Converting slide decks into searchable technical blog posts.
- Extracting financial data from multi-page PDF tables into structured files.
- Transforming scanned paper manuals into Obsidian-ready knowledge bases.
Known limitations
Large files must be processed in page ranges. Requires specific Python libraries to be installed in the environment. Merged table cells may revert to HTML for layout fidelity.
Requires local installation of poppler-utils and Python libraries like markitdown or marker-pdf. Processing speed depends on page count and OCR requirements.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Free with an Agensi account
Permissions
No special permissions declared or detected