smb-router

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFE
Full Analysis
  • [Intent Classification]: The skill interprets user natural language to identify business needs and maps them to a specific set of internal slash commands. This design facilitates efficient navigation across various business capabilities.
  • [Connector Awareness]: The instructions include logic to verify the presence of required service integrations (such as QuickBooks or HubSpot) before a command is recommended. This check ensures that the agent does not attempt to use tools that are not properly configured.
  • [User Confirmation Mechanism]: A human-in-the-loop approach is implemented by requiring the agent to ask for user permission before triggering any routed commands, ensuring the user remains in control of the agent's actions.
  • [Indirect Prompt Injection Surface]: The skill processes user input and retrieves data from session memory (ingestion points) to determine which commands to trigger. The skill addresses potential risks by requiring explicit user confirmation (boundary markers) before execution and verifying the status of required service connectors. Its capabilities are limited to triggering a predefined set of business-oriented slash commands (capability inventory) documented in the routing table, while explicit sanitization of input text is absent.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 05:33 PM