pdf-extraction-guide

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized behaviors were detected in the provided documentation or code snippets. The skill serves as an educational resource for document processing tasks.\n- [DATA_EXFILTRATION]: The skill includes code to send PDF data to a GROBID service via an HTTP POST request. By default, this is directed to a local instance (localhost:8070), which is a standard workflow for this tool and does not constitute unauthorized data exfiltration.\n- [COMMAND_EXECUTION]: The guide provides shell command examples for using well-known utilities like Pandoc and Marker. These are standard tools for document format conversion and are presented for legitimate instructional purposes.\n- [SAFE]: The skill describes processes for ingesting external PDF data, which inherently creates a surface for indirect prompt injection if the extracted text is subsequently passed to an AI model without sanitization. This is a general risk associated with the task domain and not a specific vulnerability in the skill's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 03:10 PM
Security Audit — agent-trust-hub — pdf-extraction-guide