The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/fill_form.py executes system utilities like textutil (macOS) or libreoffice/soffice to convert legacy .doc files to the .docx format. These operations are performed using subprocess.run with a list of arguments, which is the recommended practice for preventing shell command injection.
[EXTERNAL_DOWNLOADS]: The skill documentation instructs the user to install the python-docx library. This is a well-known, legitimate open-source package used for programmatically creating and updating Microsoft Word files.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by reading content from external Word documents to identify form fields.
Ingestion points: The scan_fields function in scripts/fill_form.py reads text content from tables and paragraphs in user-supplied .docx files.
Boundary markers: There are no specific boundary markers or instructions to the agent to ignore potentially malicious text found within the document content.
Capability inventory: The skill allows the agent to execute the Python script to read/write files and call system conversion tools via subprocess.run (found in scripts/fill_form.py).
Sanitization: Content read from the document is normalized for label matching but is not otherwise sanitized before being presented to the agent in the Scan phase, which could allow a malicious document to influence the agent's next steps.

lovstudio-fill-form