pdf-reader
Pass
Audited by Gen Agent Trust Hub on Jul 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions and execution commands utilize the
uvtool to dynamically fetch thepymupdflibrary.PyMuPDFis a well-known, reputable open-source library for PDF manipulation and analysis. - [PROMPT_INJECTION]: The skill processes content from external, untrusted PDF files including text, metadata, and visual renders. This constitutes an indirect prompt injection surface where instructions embedded in a PDF could attempt to influence agent behavior.
- Ingestion points: Content is read into the agent context via
pdf_info.py,pdf_extract.py, andpdf_search.py. - Boundary markers: The scripts do not wrap extracted text in specific delimiters or safety headers before outputting to the agent.
- Capability inventory: The skill has the ability to read files from the filesystem and write rendered images to the system's temporary directory (
/tmp). - Sanitization: No sanitization or filtering of the PDF content is performed prior to display.
Audit Metadata