The Agent Skills Directory

[SAFE]: The skill's instructions and implementation were thoroughly reviewed and found to be free of malicious patterns. All operations are limited to the scope of document conversion.
[COMMAND_EXECUTION]: Helper scripts batch_convert.py and convert_docx.py use subprocess.run to execute the pandoc utility. This is done securely by passing arguments as a list, which avoids shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill uses the mammoth library, a well-known and trusted tool for document conversion, which is considered a safe dependency.
[PROMPT_INJECTION]: The skill processes untrusted Word documents, which introduces an inherent surface for indirect prompt injection. If the converted Markdown contains malicious instructions and is subsequently read by an AI agent, it could influence the agent's actions. This is a common risk for tools that process external content.
Ingestion points: Word documents (.docx) processed in the conversion scripts.
Boundary markers: None identified; the tool converts content as-is.
Capability inventory: Local command execution (pandoc) and file system writes.
Sanitization: None; the tool's objective is to maintain original document content fidelity.

docx-to-md