The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill requires external Python packages (python-docx, pdf2image) and system tools (libreoffice, poppler). While these are standard tools for document processing, they involve downloads from public registries (PyPI, Homebrew, APT).
[COMMAND_EXECUTION] (LOW): The skill executes system commands via subprocess.run to call soffice and pdftoppm. The inputs to these commands (file paths) are derived from the input_path argument. While the script uses abspath and expanduser, there is a minor risk of command injection if the agent processes maliciously crafted filenames without further sanitization by the shell environment.
[REMOTE_CODE_EXECUTION] (SAFE): No patterns of piping remote scripts to shells (e.g., curl | bash) were detected.
[DATA_EXFILTRATION] (SAFE): Network operations are not used for exfiltration. Activity is confined to local file transformation.
[PROMPT_INJECTION] (SAFE): The SKILL.md contains instructional content for the agent but does not include markers or patterns typical of jailbreaks or system prompt extraction.

doc