ocr-and-documents
Pass
Audited by Gen Agent Trust Hub on Mar 5, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill incorporates the marker-pdf library for high-quality OCR, which involves a one-time download of approximately 2.5GB of machine learning models from HuggingFace to the local cache. HuggingFace is a widely recognized and trusted platform for AI resources.
- [EXTERNAL_DOWNLOADS]: The instructions reference the web_extract tool for processing remote documents, specifically citing Arxiv as a primary example of a trusted source for academic documents.
- [COMMAND_EXECUTION]: The utility scripts extract_marker.py and extract_pymupdf.py are designed for document parsing using established libraries. They perform safe file operations and data extraction without executing arbitrary shell commands or unverified remote scripts.
- [PROMPT_INJECTION]: The provided instructions are strictly functional for document processing and do not include patterns typical of prompt injection or safety bypass attempts.
Audit Metadata