skills/xberg-io/xberg/pdf/Gen Agent Trust Hub

pdf

Pass

Audited by Gen Agent Trust Hub on May 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides documentation and code snippets for processing PDF files using legitimate, industry-standard libraries such as pypdf, pdfplumber, and reportlab. No evidence of malicious behavior, obfuscation, or unauthorized data access was found.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides templates for extracting text from external PDF files. This creates an inherent surface for indirect prompt injection where instructions embedded within a document could attempt to influence agent behavior. \n
  • Ingestion points: SKILL.md (templates using pypdf.PdfReader, pdfplumber.open, and pytesseract.image_to_string)\n
  • Boundary markers: Absent in the provided code templates.\n
  • Capability inventory: File system writes (outputting merged/split PDFs) and command-line execution (via poppler-utils and qpdf).\n
  • Sanitization: None provided in the templates; the skill assumes processing of trusted or user-provided documents.
Audit Metadata
Risk Level
SAFE
Analyzed
May 21, 2026, 08:51 PM
Security Audit — agent-trust-hub — pdf