The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-supplied technical descriptions to generate documents and perform searches, which establishes a surface for indirect prompt injection.
Ingestion points: Technical descriptions are ingested through scripts/generate_disclosure.py and scripts/patent_search.py via command-line arguments and file inputs.
Boundary markers: Technical content is delimited using Markdown headers in the generated technical disclosure templates.
Capability inventory: The skill performs network reads for patent data and local file writes for generated documents.
Sanitization: Search queries are URL-encoded in scripts/patent_search.py to ensure web request safety.
[EXTERNAL_DOWNLOADS]: The scripts/patent_search.py script communicates with legitimate external patent databases and academic search engines, including Google Patents, Lens.org, Espacenet, and CNIPA, to retrieve search results based on user-provided queries.
[COMMAND_EXECUTION]: The skill utilizes local Python scripts to execute its core functionality. These scripts are invoked by the agent as documented and rely exclusively on the Python standard library without calling dangerous external binaries.

patent-assistant