openviking-rule-search

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to retrieve and follow rules from an external tool (rule_search or its variants), which represents a classic indirect prompt injection surface where the agent's behavior is guided by externally provided content.\n
  • Ingestion points: The agent reads data from the rule_search tool output to determine constraints for its current task.\n
  • Boundary markers: The skill includes instructions to read results in "bounded sections" and handle oversized responses, providing some degree of structural awareness.\n
  • Capability inventory: The skill instructs the agent to "adjust the planned action or response to comply with all applicable Rules," which is a high-influence capability. However, the skill does not define dangerous local capabilities such as direct file system writes or unauthorized network requests.\n
  • Sanitization: There are no instructions for sanitizing or validating the content of the rules before they are adopted as constraints.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 08:10 AM