The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system commands to perform its functions:
scripts/office/soffice.py executes gcc to compile a technical shim (lo_socket_shim.c) at runtime to allow LibreOffice to function in restricted environments.
scripts/accept_changes.py invokes soffice (LibreOffice) to programmatically accept tracked changes in documents.
scripts/office/validators/redlining.py uses git diff to identify and validate changes made to document text content.
SKILL.md documents and triggers the use of pandoc for text extraction and pdftoppm (Poppler) for image conversion.
[EXTERNAL_DOWNLOADS]: The instructions in SKILL.md direct the agent to install the docx package globally using npm install -g docx for generating new documents.
[DYNAMIC_EXECUTION]: The script scripts/office/soffice.py contains an embedded C source string that is compiled and loaded via LD_PRELOAD at runtime. While this is used as a workaround for socket restrictions in sandboxed environments, it involves the dynamic generation and loading of executable code.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it processes untrusted Office documents:
Ingestion points: Document files are processed by scripts/office/unpack.py and pandoc (via SKILL.md).
Boundary markers: None identified in instructions for separating document content from agent instructions.
Capability inventory: The skill has significant capabilities including file writing, network/socket shimming, and subprocess execution across multiple scripts.
Sanitization: The skill uses defusedxml for XML parsing and provides an extensive validation system to ensure XML structure conforms to Office Open XML schemas.

docx