The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external commands and recursive script calls to perform its primary functions.
The scripts/fill_template.py script spawns sub-processes of itself using subprocess.run to facilitate batch processing of document elements.
The scripts/layout_gate.py and scripts/ole2_to_docx.py scripts invoke external binaries (soffice from LibreOffice) to convert legal documents between DOC, DOCX, and PDF formats.
These operations are used purely for document processing and layout verification.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content which creates a surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent context through legal documents (Markdown or Docx) read by scripts/extract_from_markdown.py and the main Agent extraction logic.
Boundary markers: The skill does not implement specific delimiters or warnings to instruct the LLM to ignore potentially malicious instructions embedded in the input documents.
Capability inventory: The skill possesses capabilities to execute shell commands (via subprocess) and perform local filesystem operations.
Sanitization: There is no evidence of sanitization or filtering of the input text before it is used for data extraction or template filling.

elements-complaint-generator