The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py contains logic to dynamically write C source code to the temporary directory and compile it into a shared object file using gcc at runtime. It then utilizes the LD_PRELOAD environment variable to inject this library into the soffice binary. While used as a workaround for socket restrictions in sandboxed environments, runtime compilation and library injection are highly privileged operations.
[COMMAND_EXECUTION]: The scripts/office/unpack.py and scripts/office/validate.py scripts utilize zipfile.extractall() to extract content from Word documents. This method is vulnerable to Zip Slip attacks because it does not validate that extracted file paths remain within the intended directory, potentially allowing a malicious document to overwrite arbitrary system files via path traversal.
[COMMAND_EXECUTION]: Multiple scripts invoke system tools such as soffice, gcc, git, and pandoc using subprocess.run. These executions are fundamental to the skill's document conversion and validation capabilities but involve interaction with the underlying operating system that requires careful oversight.
[SAFE]: Document XML files are parsed using the defusedxml library, which protects the agent and host system from common XML-based vulnerabilities including Billion Laughs attacks and XML External Entity (XXE) injection.

docx