web-search

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for tools that ingest external, untrusted content from the internet, which can contain hidden instructions targeting the AI agent.
  • Ingestion points: Content extracted from URLs via the tavily/extract and exa/extract apps mentioned in the documentation.
  • Boundary markers: Workflows use delimiters like <search-results> for data interpolation, but do not provide explicit instructions to the agent to disregard instructions within these blocks.
  • Capability inventory: The skill operates within a restricted bash environment limited to the belt CLI command.
  • Sanitization: The instructions do not specify any sanitization or validation of the retrieved web content before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 09:05 PM