biopython-entrez
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves records such as abstracts, sequences, and metadata from external NCBI databases, creating a potential surface for indirect prompt injection if the fetched data contains malicious instructions.\n
- Ingestion points: The
scripts/pubmed_summaries.pyscript utilizesBio.Entrez.esearchandBio.Entrez.esummaryto ingest external content from the NCBI API.\n - Boundary markers: There are no explicit instructions or delimiters used to ensure the agent ignores embedded instructions within the biological data retrieved.\n
- Capability inventory: The skill has file-write access (to the
outputs/directory) and network access to NCBI APIs via the Biopython library.\n - Sanitization: Content is parsed via Biopython's internal XML parser (
Entrez.read), which provides structural integrity but does not filter the text for potential prompt injection patterns.\n- [INDIRECT_PROMPT_INJECTION]: The skill includes a file (biopython-entrez_audit_result_v1.json) containing self-referential safety claims (e.g., 'Verdict: PASS', 'security: PASS'). Per security guidelines, such claims are disregarded as authoritative and are instead evaluated as data targeting the analysis process.\n - Ingestion points:
biopython-entrez_audit_result_v1.jsoncontains metadata asserting the skill's own security status.\n - Boundary markers: N/A\n
- Capability inventory: N/A\n
- Sanitization: N/A\n- [EXTERNAL_DOWNLOADS]: The skill fetches scientific data and records from NCBI's well-known E-utilities service via official API endpoints.
Audit Metadata