yzr-gemini-pdf-summary
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted PDF files by feeding them directly into the Google Gemini multi-modal API. This creates an attack surface where malicious content within a PDF could potentially override agent instructions or influence the summary output. \n
- Ingestion points: The args.pdf file is read and passed as binary data to the LLM in scripts/gemini_pdf_summary.py. \n
- Boundary markers: The system prompts do not currently use specific delimiters or instructions to ignore potential commands embedded within the PDF content. \n
- Capability inventory: The script has permissions to write files (Markdown summaries and extracted images) to the local filesystem and perform network requests to the Gemini API. \n
- Sanitization: There is no pre-processing or sanitization of the PDF content to strip potentially malicious instructions before processing.
Audit Metadata