biorxiv-database
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by ingesting untrusted data from an external research database. 1. Ingestion points: Fetches preprint titles, abstracts, and metadata from the bioRxiv API via 'scripts/biorxiv_search.py'. 2. Boundary markers: The instructions lack delimiters or warnings to ignore instructions embedded within the retrieved content. 3. Capability inventory: The skill can write files to the local system, including search results in JSON format and PDF downloads. 4. Sanitization: No sanitization or filtering logic is described for the retrieved external content before it is processed by the agent.
- [EXTERNAL_DOWNLOADS]: Connects to the official bioRxiv.org domain to retrieve preprint information and download full-text PDFs. It also recommends the installation of the standard 'requests' Python library.
- [COMMAND_EXECUTION]: Provides instructions for executing local Python scripts ('scripts/biorxiv_search.py') to perform search queries and data management tasks.
Audit Metadata