The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/generate_word.py script executes system commands (open on macOS, xdg-open on Linux, and os.startfile on Windows) to automatically open the folder containing the generated report. This is a functional feature designed to improve user experience by showing the output file immediately.
[EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md specify that the python-docx library should be installed at runtime if not present. This is a well-known, standard library for manipulating Word documents and is considered a safe dependency for this use case.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external patent documents provided by the user. This creates an attack surface for indirect prompt injection where instructions could be hidden in patent text. However, the skill implements a strict data contract and validation logic (seen in scripts/review_validation.py and scripts/generate_word.py) that requires structured JSON input and specific evidence chains, which helps mitigate the risk of the agent following instructions embedded within the analyzed content.

patent-quality-review-pro