citation-management

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The documentation identifies several standard Python dependencies required for the skill's operation, including requests, biopython, and scholarly. These are well-known libraries used for HTTP communication and academic data retrieval.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external academic search engines and repositories. This represents a potential surface for indirect prompt injection if paper metadata were to contain malicious instructions, though this is a characteristic of the research task itself.
  • Ingestion points: Metadata retrieved from Google Scholar, PubMed, and arXiv APIs, as well as local BibTeX files.
  • Boundary markers: Not explicitly defined in the provided instructional text.
  • Capability inventory: The skill scripts perform network requests to academic services and write resulting citation data to the local file system.
  • Sanitization: Includes logic for protecting capitalization in BibTeX titles and standardizing author names, ensuring data integrity for citation purposes.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 05:11 AM
Security Audit — agent-trust-hub — citation-management