The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill is designed for academic presentation generation and uses standard PDF and image processing libraries. It does not perform any unauthorized network operations or exfiltrate sensitive data.\n- [COMMAND_EXECUTION]: Static analysis identified a call to subprocess.run in src/sci_html/integrations/figure/utils.py. Analysis confirms this is a benign utility function that checks for the presence and version of the Tesseract OCR engine on the local system using a controlled command line.\n- [SAFE]: The use of __import__ in src/sci_html/integrations/figure/utils.py was investigated. It is used exclusively to verify that required dependencies (such as numpy and Pillow) are installed, using a hardcoded list of package names rather than user-provided input.\n- [PROMPT_INJECTION]: The skill handles untrusted user data from PDFs and Markdown files. Risks of indirect prompt injection are mitigated by structured data extraction logic and strict HTML escaping in the renderer (src/sci_html/renderer.py) before the final presentation is generated.

sci-html