biorxiv-search

Pass

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a local Bash script (scripts/search) and a Node.js CLI tool (scripts/search.mjs) to execute search queries and handle API configuration. It includes a specific command to locate these scripts within the environment's plugin cache.
  • [EXTERNAL_DOWNLOADS]: The skill communicates with the external Valyu API (api.valyu.ai) to retrieve biological preprint data and full-text content. This is a legitimate functional requirement for the search service.
  • [PROMPT_INJECTION]: The skill processes untrusted external data (preprint content) from bioRxiv, which represents an indirect prompt injection surface.
  • Ingestion points: Article content and titles fetched from api.valyu.ai via bioRxiv.
  • Boundary markers: Absent; there are no specific delimiters or instructions to ignore commands within the fetched research papers.
  • Capability inventory: The skill allows for local script execution and processing of search results via jq.
  • Sanitization: No explicit sanitization or filtering of the fetched content is described in the skill instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 6, 2026, 11:29 PM
Security Audit — agent-trust-hub — biorxiv-search