OpenSkillIndex← back to search
claude-code

Oh-my-agent oma-pdf

install
source · Clone the upstream repo
git clone https://github.com/first-fluke/oh-my-agent
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/first-fluke/oh-my-agent "$T" && mkdir -p ~/.claude/skills && cp -r "$T/.agents/skills/oma-pdf" ~/.claude/skills/first-fluke-oh-my-agent-oma-pdf && rm -rf "$T"
manifest: .agents/skills/oma-pdf/SKILL.md
source content

PDF Skill - PDF to Markdown Conversion

When to use

  • Converting PDF documents to Markdown for LLM context or RAG
  • Extracting structured content (tables, headings, lists) from PDFs
  • Preparing PDF data for AI consumption
  • User says "convert this PDF", "parse PDF", "PDF to markdown", "read this PDF"

When NOT to use

  • Generating or creating PDFs -> use appropriate document tools
  • Editing existing PDFs -> out of scope
  • Simple file reading of already-text files -> use Read tool directly

Core Rules

  1. Use 
    uvx opendataloader-pdf
    to run — no installation required
  2. Default output format is Markdown
  3. If no output directory specified, output to the same directory as the input PDF
  4. Preserve document structure: headings, tables, lists, images
  5. For scanned PDFs, use hybrid mode with OCR
  6. Always run 
    uvx mdformat
    on the output to normalize Markdown formatting
  7. Validate the output Markdown is readable and well-structured
  8. Report any conversion issues (missing tables, garbled text) to the user

How to Execute

Follow 

resources/execution-protocol.md
step by step.

Quick Reference

Basic conversion (single file)

uvx opendataloader-pdf input.pdf

Specify output directory

uvx opendataloader-pdf input.pdf --output-dir ./output/

Multiple files or folder

uvx opendataloader-pdf file1.pdf file2.pdf folder/

With OCR (scanned PDFs)

Requires hybrid mode server:

uvx opendataloader-pdf-hybrid --port 5002 --force-ocr --ocr-lang "ko,en"
uvx opendataloader-pdf --hybrid docling-fast input.pdf

With image extraction (embedded base64)

uvx opendataloader-pdf input.pdf --image-output embedded --image-format png

With Tagged PDF structure

uvx opendataloader-pdf input.pdf --use-struct-tree

Output Formats

FormatFlagUse case
Markdown
--format markdown
Default. Clean text for LLM/RAG
JSON
--format json
Structured data with bounding boxes
HTML
--format html
Web display
Text
--format text
Plain text extraction
Combined
--format markdown,json
Multiple formats at once

Configuration

Project-specific settings: 

config/pdf-config.yaml

Troubleshooting

IssueSolution
Garbled text in outputTry 
--use-struct-tree
for Tagged PDFs
Scanned PDF (no text layer)Use hybrid mode with 
--force-ocr
Tables not extracted properlyUse hybrid mode for complex/borderless tables
Non-English PDFAdd 
--ocr-lang
with appropriate language codes
Large PDF (100+ pages)Process in page ranges or use batch mode
Formula not extractedUse hybrid mode with 
--enrich-formula

References

  • Execution steps: 
    resources/execution-protocol.md
  • Configuration: 
    config/pdf-config.yaml
  • Context loading: 
    ../_shared/core/context-loading.md
  • Quality principles: 
    ../_shared/core/quality-principles.md