pdfkit

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from external PDF files, creating a potential surface for indirect prompt injection if malicious content within a PDF is processed and used to influence agent behavior. \n
  • Ingestion points: The skill demonstrates loading PDF documents from URLs and raw data using PDFDocument(url:) and PDFDocument(data:). \n
  • Boundary markers: The provided code snippets do not utilize delimiters or specific instructions to the agent to ignore embedded commands within the PDF content. \n
  • Capability inventory: The skill includes methods for writing files to the system via document.write(to:) and opening external network URLs using UIApplication.shared.open(url). \n
  • Sanitization: No sanitization, validation, or filtering of the text extracted from PDF documents is demonstrated before it is used in the application context.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 11:25 PM
Security Audit — agent-trust-hub — pdfkit