pubmed-database
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. All external communications are directed to the trusted and well-known NCBI domain (eutils.ncbi.nlm.nih.gov).\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it processes third-party scientific literature and abstracts fetched from the PubMed API.\n
- Ingestion points: PubMed API responses (efetch.fcgi, esummary.fcgi) as documented in SKILL.md and references/api_reference.md.\n
- Boundary markers: None identified in the instructional snippets or code examples.\n
- Capability inventory: The agent can perform network requests to official NCBI endpoints.\n
- Sanitization: No specific sanitization or filtering of retrieved content is described.\n- [SAFE]: The technical support email address provided in SKILL.md and references/api_reference.md (vog.hin.mln.ibcn@seitilitue) is obfuscated using string reversal as an anti-spam measure. It decodes to the legitimate contact eutilities@ncbi.nlm.nih.gov.
Audit Metadata