The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it extracts and processes content from untrusted external Word documents (.docx).
Ingestion points: The scripts/parse_docx.py script reads the contents of user-provided Word files using the python-docx, markitdown, and docx2txt libraries.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the extracted markdown content as untrusted or to ignore any embedded instructions found within the documents.
Capability inventory: The script performs file system operations including directory creation (mkdir) and writing extracted text, tables, and images to files. The resulting markdown artifacts are intended for subsequent consumption by the AI agent.
Sanitization: No sanitization or filtering is performed on the text extracted from the document before it is converted to markdown. If a processed document contains malicious instructions aimed at the AI agent, the agent may follow them when it reads the parsed output.
[EXTERNAL_DOWNLOADS]: The script scripts/parse_docx.py uses the uv package manager to manage its environment and dependencies.
It downloads and installs standard document processing libraries: python-docx, markitdown (a well-known service provided by Microsoft), and docx2txt.

docs-docx