The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill instructs the agent to analyze the contents of Word documents by converting them to Markdown using Pandoc and reading the full result. This exposes the agent to potentially malicious instructions embedded within user-provided documents.
Ingestion points: Document text extraction via pandoc as described in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when reading the extracted Markdown content.
Capability inventory: The skill includes Python scripts (pack.py, redlining.py, validate.py) that perform subprocess execution for document validation and comparison.
Sanitization: The skill mitigates common document-based threats by using defusedxml for secure XML parsing and html.escape for author metadata processing.
[COMMAND_EXECUTION]: The skill uses subprocess.run to execute well-known tools including soffice (LibreOffice), git, and pdftoppm for document conversion, differencing, and imaging. These calls use the list-based argument format without shell=True, which is a safe practice that prevents shell injection.
[SAFE]: The skill follows security best practices for handling Office Open XML (OOXML) files, specifically using specialized XML parsers designed to prevent XML External Entity (XXE) vulnerabilities.

docx