The Agent Skills Directory

[COMMAND_EXECUTION]: The bundled helper script scripts/render_docx.py and instructions in SKILL.md execute system tools including soffice (LibreOffice) and pdftoppm (Poppler) to convert and render documents. The Python script correctly uses list-based arguments with subprocess.run() and avoids shell=True, which is a security best practice to prevent command injection.
[PRIVILEGE_ESCALATION]: The SKILL.md file includes setup instructions for Ubuntu/Debian that utilize sudo apt-get install to install system dependencies. This requires administrative privileges but is a standard requirement for the identified dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill processes external .docx files, which presents a surface for indirect prompt injection if an attacker-controlled document contains malicious instructions intended to influence the agent.
Ingestion points: The skill reads file paths and content from external documents via python-docx and the rendering script in scripts/render_docx.py.
Boundary markers: The skill documentation explicitly includes an Anti-Pattern section instructing the agent not to treat external content or tool responses as trusted instructions.
Capability inventory: The skill has the ability to execute subprocess commands and write files to the local file system (e.g., in tmp/docs/ and output/doc/).
Sanitization: While the skill provides clear behavioral guidelines, the rendering script does not perform automated sanitization of the text content extracted from documents.
[EXTERNAL_DOWNLOADS]: The skill workflow involves downloading and installing well-known Python packages (python-docx, pdf2image) and system utilities from official package registries.

doc