web-search-advanced-research-paper

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • Indirect Prompt Injection (LOW): This skill defines a workflow that ingests untrusted external data via the web_search_advanced_exa tool, which creates a surface for indirect prompt injection.
  • Ingestion points: Search results from academic databases (arXiv, OpenReview, etc.) are brought into the agent context.
  • Boundary markers: The instructions do not define delimiters (such as XML tags or triple backticks) to isolate the untrusted search results from the agent's core instructions.
  • Capability inventory: The skill itself is purely instructional and does not provide script execution, file system access, or network-writing capabilities.
  • Sanitization: There are no instructions for sanitizing the output or explicitly telling the agent to ignore instructions embedded in the search results.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:28 PM