pdf-reading
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (PDF files) from disk, which creates a surface for indirect prompt injection where malicious instructions hidden in the PDF content (metadata, text, or OCR) could influence the agent's behavior.
- Ingestion points: Processes files from disk using
pdfinfo,pdftotext,pdfplumber, andpypdf(found inSKILL.md). - Boundary markers: None specified in the instructions to separate extracted document text from agent instructions.
- Capability inventory: The skill utilizes shell command execution (
pdftotext,pdfimages,pdfdetach,pdftoppm) and Python script execution (pypdf,pdfplumber,fitz). - Sanitization: No explicit sanitization or filtering of the extracted text or metadata is described, although a basic filename sanitization (
os.path.basename) is mentioned for attachments.
Audit Metadata