pdf-to-markdown
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
subprocess.runto interface with thexbergutility. The implementation passes arguments as a list and specifically avoidsshell=True, which prevents shell injection attacks from malformed filenames or parameters. - [INDIRECT_PROMPT_INJECTION]: As the skill ingests and processes external PDF documents for conversion, it naturally has an indirect prompt injection surface. However, the
scripts/convert_pdf.pywrapper mitigates this risk by performing structural validation, including verifying contiguous page metadata and reconciling page markers before the content is presented to the agent. - [EXTERNAL_DOWNLOADS]: The skill notes that enabling Korean OCR via PaddleOCR may trigger the download of necessary models. The instructions explicitly mandate that the agent must notify the user of potential network usage and model storage requirements before proceeding.
Audit Metadata