paddleocr-text-recognition

Pass

Audited by Gen Agent Trust Hub on May 6, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/ocr_caller.py and scripts/smoke_test.py) to manage OCR tasks and perform configuration checks.
  • [EXTERNAL_DOWNLOADS]: The skill makes outbound network requests to an external API service, as defined by the user-provided PADDLEOCR_OCR_API_URL environment variable.
  • [DATA_EXFILTRATION]: User-provided files and document URLs are transmitted to the external OCR service for processing. This data movement is the intended function of the skill but involves sending document content to a remote server.
  • [PROMPT_INJECTION]: The skill extracts text from untrusted documents and instructs the agent to display the results in full. This creates an indirect prompt injection surface where malicious text embedded within an image or PDF could attempt to influence the agent's behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
May 6, 2026, 11:39 AM