hebrew-ocr-forms
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes text extracted via OCR from user-provided images, which creates an attack surface where malicious text within a document could attempt to influence the agent's behavior.\n
- Ingestion points: The scripts
scripts/extract_form_fields.pyandscripts/tabu_to_spreadsheet.pyingest data from processed images and OCR text.\n - Boundary markers: The skill relies on regex patterns for field extraction rather than explicit prompt delimiters to separate processed text from agent instructions.\n
- Capability inventory: The scripts perform local file operations and image processing but do not possess network capabilities.\n
- Sanitization: Data is filtered using regex patterns, and Israeli IDs are validated using a standard check-digit algorithm.\n- [EXTERNAL_DOWNLOADS]: The skill provides instructions and links for downloading official resources from trusted and well-known services.\n
- References the official Tesseract OCR GitHub repository for downloading Hebrew language trained data.\n
- Links to official technical documentation from Google Cloud, AWS, Microsoft, and the Israeli government.\n- [COMMAND_EXECUTION]: The skill instructions involve executing local Python scripts that utilize the Tesseract OCR engine. This command execution is necessary for the skill's primary function of image processing and text extraction.
Audit Metadata