The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network operations to retrieve paper metadata from well-known academic services, including export.arxiv.org, api.clarivate.com (Web of Science), and eutils.ncbi.nlm.nih.gov (PubMed). These requests are limited to the user's search query and do not transmit sensitive local data.\n- [COMMAND_EXECUTION]: The tool executes the sci_search.py script to manage search requests, process results, and handle library persistence. This is standard behavior for an agent skill of this type.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes and displays data from external academic sources.\n
Ingestion points: External data is fetched from the arXiv, PubMed, and Web of Science APIs in sci_search.py.\n
Boundary markers: While search results are structured using Markdown headers and bullet points, the skill does not provide explicit instructions for the agent to ignore instructions embedded within the paper metadata.\n
Capability inventory: The skill writes to a local library.json file for caching and supports writing search reports to user-specified file paths.\n
Sanitization: Content such as paper titles and abstracts are utilized directly without specific sanitization against prompt injection attacks.

sci-search