pubmed-database
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEOBFUSCATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides documentation and programmatic access examples for the PubMed database through the official NCBI E-utilities REST API (eutils.ncbi.nlm.nih.gov). This is a legitimate and well-known service provided by the National Center for Biotechnology Information.
- [SAFE]: The skill recommends using standard and well-known Python libraries such as
requestsandbiopythonfor API interaction. - [OBFUSCATION]: A technical support email address in
SKILL.mdis provided in reverse format (vog.hin.mln.ibcn@seitilitueforeutilities@ncbi.nlm.nih.gov). This is interpreted as a benign anti-spam measure for a public contact address and does not hide malicious functionality. - [INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection by ingesting external data from the PubMed database.
- Ingestion points: Literature metadata, abstracts, and full records are retrieved from the NCBI E-utilities API (
esearch.fcgi,efetch.fcgi,esummary.fcgi). - Boundary markers: No explicit boundary markers or instructions to ignore embedded prompts are defined for the fetched content.
- Capability inventory: Capabilities are restricted to network read operations and response parsing. The skill does not implement high-risk tools such as subprocess execution, dynamic code evaluation, or file system modifications.
- Sanitization: The skill involves standard parsing of structured JSON, XML, or plain text responses from the official API.
Audit Metadata