skills/pascalorg/skills/image-to-text/Gen Agent Trust Hub

image-to-text

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The shell script (scripts/image-to-text.sh) performs an automated npm install from the official npm registry to initialize the local environment with required packages.
  • [EXTERNAL_DOWNLOADS]: The Tesseract.js library is designed to fetch necessary OCR language training models (e.g., eng.traineddata) from its official GitHub repository during the first execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted image files via the imagePath argument in scripts/image-to-text.js. While the extracted text is returned as structured JSON, no specific sanitization is applied. This creates a surface where text-based instructions inside processed images could potentially be interpreted by an agent, which is a common risk for OCR tools.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:18 AM
Security Audit — agent-trust-hub — image-to-text