ocr-and-documents

Pass

Audited by Gen Agent Trust Hub on Jul 1, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the agent to install official Python packages including pymupdf, marker-pdf, and python-docx. It also notes that marker-pdf downloads model weights from Hugging Face, which is a well-known and trusted service.
  • [COMMAND_EXECUTION]: The skill utilizes local shell commands to run helper scripts (scripts/extract_pymupdf.py, scripts/extract_marker.py) and perform PDF operations such as splitting and merging via Python snippets. These operations are standard for the skill's stated purpose of document processing.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface as it processes untrusted external data (PDFs, scans, and web content) and converts it into text for the agent.
  • Ingestion points: Untrusted data enters the agent context through web_extract and local extraction scripts.
  • Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in the processed output.
  • Capability inventory: The skill uses subprocess calls for package installation and file system access for reading and writing documents.
  • Sanitization: There is no evidence of sanitization or filtering of the extracted content before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 1, 2026, 09:48 AM