pdf-to-markdown

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • External Downloads (LOW): The skill installs several Python packages via uv pip install and downloads large AI model files from Hugging Face during its first run. Hugging Face is recognized as a trusted organization, which minimizes the risk associated with these downloads.
  • Evidence: Installation of pymupdf, docling, and docling-core in SKILL.md; model presence check using huggingface_hub in scripts/extractor.py.
  • Prompt Injection (LOW): The skill is susceptible to indirect prompt injection because it processes user-provided PDF files that may contain malicious instructions designed to influence the agent's logic during content analysis.
  • Ingestion points: PDF file reading and processing in scripts/extractor.py.
  • Boundary markers: The output includes a YAML metadata block, but lacks explicit delimiters or instructions to ignore embedded commands within the extracted text.
  • Capability inventory: The skill can read/write local files and execute Python scripts within a dedicated virtual environment.
  • Sanitization: No sanitization or filtering of the extracted PDF text content is implemented.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:39 PM