The Agent Skills Directory

[SAFE]: No malicious patterns, hidden code, or unauthorized behaviors were detected in the skill instructions or reference documentation.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates access to genomic datasets and metadata from well-known scientific sources, specifically the official NCBI FTP and E-utilities servers (ncbi.nlm.nih.gov).\n- [PROMPT_INJECTION]: The skill ingests metadata from external genomic databases, which constitutes a surface for indirect prompt injection.\n
Ingestion points: External data is retrieved via GEOparse.get_GEO, Entrez.esummary, and Entrez.efetch in SKILL.md and references/geo_reference.md.\n
Boundary markers: No explicit delimiters are present to isolate external content from the agent's internal logic.\n
Capability inventory: The skill possesses capabilities for file system write operations (to the ./data directory) and network access to scientific services.\n
Sanitization: No specific validation or sanitization of external metadata was identified in the provided scripts.

geo-database