The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs runtime C code compilation and library injection. scripts/office/soffice.py dynamically generates a C source file in the temporary directory, compiles it into a shared object using gcc, and then uses the LD_PRELOAD environment variable to inject the resulting binary into LibreOffice subprocesses. This technique is used to shim system-level socket calls.
[COMMAND_EXECUTION]: The skill uses subprocess.run to execute multiple external binaries and system utilities. This includes running soffice for document conversion, gcc for compilation, git diff for comparing document text, and pandoc or pdftoppm for processing document content and images.
[PROMPT_INJECTION]: The skill has a significant vulnerability surface for indirect prompt injection. It extracts text and XML from untrusted user-supplied Word documents (scripts/office/unpack.py) without applying sanitization, delimiters, or boundary markers. This data is then processed in an environment where the agent has access to powerful tools like compilers and system utilities, creating a risk that instructions embedded in a document could be executed.
[EXTERNAL_DOWNLOADS]: The instructions in SKILL.md require the global installation of the docx Node.js package via npm install -g docx. The skill also depends on the presence of system-level software including LibreOffice, pandoc, and Poppler utilities.

docx