ocr

Pass

Audited by Gen Agent Trust Hub on Jun 6, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/setup_check.py script utilizes subprocess.run to check for the presence and version of local system tools including ollama and poppler (via pdftoppm). These calls use static command lists and do not incorporate untrusted user input, making them safe for environment verification.
  • [SAFE]: All network operations are directed to localhost:11434 by default to interact with the local Ollama service. This ensures that sensitive document data remains on the user's local infrastructure and is not transmitted to external third-party servers.
  • [SAFE]: The skill's dependencies (e.g., requests, Pillow, pdf2image) are well-known, standard libraries used for their intended purposes of HTTP communication, image processing, and PDF conversion.
  • [SAFE]: While the skill processes external data (images and PDFs) that could contain instructions intended for the agent (Indirect Prompt Injection), this is a standard risk for OCR tools. The skill does not implement any mechanisms that would increase this risk or bypass existing agent safety guardrails.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 6, 2026, 09:59 AM
Security Audit — agent-trust-hub — ocr