ocr-to-en-word

Fail

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The file .key contains a hardcoded APP_ID, API_SECRET, and API_KEY used for authentication with iFlytek OCR services.
  • [DATA_EXFILTRATION]: Local documents (PDFs and images) are transmitted to iFlytek's servers at cbm01.cn-huabei-1.xf-yun.com and iocr.xfyun.cn to perform OCR tasks.
  • [EXTERNAL_DOWNLOADS]: The skill downloads processed Word documents from dynamic URLs provided by the OCR service, typically hosted on iFlytek's CDN (bjcdn.openstorage.cn). Additionally, the pdf_ocr.py script allows initiating OCR tasks from arbitrary remote PDF URLs via the --pdf-url parameter.
  • [COMMAND_EXECUTION]: The skill utilizes Python scripts (ocr_translate.py, translate_to_en.py, pdf_ocr.py, image_ocr.py) that perform file system and network operations. These scripts use zipfile.ZipFile.extractall() to unpack Word documents without validating the target paths of zip members, creating a vulnerability to path traversal (Zip Slip) attacks if a malicious document is processed.
  • [COMMAND_EXECUTION]: The script ocr_translate.py dynamically modifies sys.path to import modules from a subdirectory (ocr-skill/scripts), which is a form of dynamic loading from computed paths.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 17, 2026, 09:46 AM
Security Audit — agent-trust-hub — ocr-to-en-word