The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill possesses a significant attack surface by ingesting external content from NCBI and having file-writing capabilities. * Ingestion points: Data is ingested via Entrez.read (multiple locations) and GEOparse.get_GEO (line 211). * Boundary markers: None are present to distinguish between instructions and data. * Capability inventory: The skill can write files to the local system using urllib.request.urlretrieve (line 247) and shell commands like wget (line 233). * Sanitization: There is no evidence of sanitization or filtering of the external data before it enters the agent's context.
Unverifiable Dependencies (MEDIUM): The skill documentation recommends installing biopython and GEOparse via pip. These are external packages from public repositories that are not within the explicitly trusted provider scope.
External Downloads (MEDIUM): The skill is designed to perform network operations to ncbi.nlm.nih.gov and ftp.ncbi.nlm.nih.gov to retrieve biological datasets and metadata.
Command Execution (LOW): The skill includes bash snippets for using wget to download data, which involves interacting with the system shell.

bio-geo-data