The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple command-line utilities including pandoc, soffice, pdftoppm, and git to perform document conversions, rendering, and character-level comparisons. It also utilizes several internal Python scripts for XML manipulation and validation.
[COMMAND_EXECUTION]: Instructions explicitly direct the agent to use sudo apt-get install for system dependencies, which involves executing commands with administrative privileges.
[EXTERNAL_DOWNLOADS]: The skill performs runtime installation of software packages via npm, pip, and apt-get from external repositories.
[PROMPT_INJECTION]: An indirect prompt injection surface exists when processing external document files.
Ingestion points: Untrusted content enters the agent context via Markdown generated by pandoc and raw XML from unpack.py.
Boundary markers: Absent; document content is interpolated into the agent context without the use of delimiters or protective instructions.
Capability inventory: Subprocess execution in pack.py and redlining.py, and file system write access in document.py.
Sanitization: Absent; the skill does not escape or filter document content before it is displayed to the agent for analysis.

docx