paddleocr-doc-parsing

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [SAFE]: No high-severity security vulnerabilities or malicious patterns were detected in the skill's code or instructions.
  • [EXTERNAL_DOWNLOADS]: The smoke test utility performs a download of a sample image from a well-known vendor repository (Baidu/PaddlePaddle) to verify API connectivity and configuration.
  • [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection when parsing external documents. \n
  • Ingestion points: Untrusted content from files or URLs is processed by the scripts/vl_caller.py script. \n
  • Boundary markers: The prompt template in SKILL.md does not utilize strong delimiters or explicit 'ignore instructions' warnings for the extracted text. \n
  • Capability inventory: The agent has the ability to execute local helper scripts and communicate with the configured PaddleOCR API endpoint. \n
  • Sanitization: No sanitization or filtering is performed on the text extracted from documents before it is displayed to the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 05:55 PM