search-lit

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several bundled scripts (pubmed_eutils.sh, parse_pubmed.py, check_doi_record_match.py, snowball.py) to facilitate communication with external medical databases and process bibliographic data. These scripts are implemented using standard Bash and Python and perform specific, legitimate functions.
  • [EXTERNAL_DOWNLOADS]: The skill makes network requests to PubMed (NCBI), Crossref, and Semantic Scholar APIs to retrieve article metadata. These are well-known and authoritative services for medical research. Requests are performed using curl and Python's urllib library.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external APIs, such as research paper titles and abstracts. While this provides a theoretical surface for injection, the skill mitigates risk by parsing the data into structured formats (BibTeX, JSON, XML) and presenting it in a descriptive bibliographic context.
  • [CREDENTIALS_UNSAFE]: The skill references an optional NCBI_API_KEY for increased rate limits, which is correctly managed via environment variables rather than being hardcoded. This aligns with secure credential handling practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:59 AM
Security Audit — agent-trust-hub — search-lit