The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied Markdown documents which could contain malicious instructions or hidden payloads. There are no explicit instructions to the AI agent to ignore or delimit instructions found within the source document, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads markdown files via the scripts/md2docx.py script provided in the command-line arguments.
Boundary markers: Absent; no specific delimiters or warnings are used to isolate document content from agent instructions in the prompt interpolation.
Capability inventory: The script can read local files (for image embedding), write to the filesystem (to save the DOCX), and perform network operations (to download remote images).
Sanitization: YAML frontmatter is stripped from the markdown, but no further validation or sanitization of markdown content or remote URLs is performed.
[EXTERNAL_DOWNLOADS]: The scripts/md2docx.py script uses urllib.request.urlretrieve to automatically download images from remote URLs found in the Markdown source. This is a functional requirement for image embedding but allows the skill to initiate network connections to potentially untrusted external servers based on user-provided content.
[COMMAND_EXECUTION]: The skill requires the execution of a Python script (scripts/md2docx.py) and standard package installation (pip install python-docx). These are documented and necessary steps for the skill's primary function of document conversion.

lov-any2docx