hive-query

Pass

Audited by Gen Agent Trust Hub on Jul 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill follows recommended patterns for Model Context Protocol (MCP) tool integration. It uses a structured workflow for intent routing, schema discovery, and bounded API invocation.
  • [COMMAND_EXECUTION]: There are no patterns suggesting arbitrary command execution. The skill exclusively uses defined meta-tools (search_tools, get_api_endpoint_schema, invoke_api_endpoint) to interact with the Hive infrastructure.
  • [DATA_EXFILTRATION]: The skill is designed to retrieve public blockchain and market data. It does not attempt to access sensitive local files, credentials, or environment variables.
  • [PROMPT_INJECTION]: The skill ingests data from external blockchain providers, creating an indirect prompt injection surface.
  • Ingestion points: Data enters the agent's context from external APIs (e.g., CoinGecko, DeFiLlama) via the invoke_api_endpoint tool.
  • Boundary markers: The instructions provide a "Report template" for the final answer but do not define specific delimiters to isolate potentially untrusted tool output from the agent's instructions.
  • Capability inventory: The agent can search for tools, inspect API schemas, and execute calls to various crypto data providers.
  • Sanitization: The instructions promote the use of "bounded arguments" such as limits and time windows to control the data volume, although they do not specify content-level sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 24, 2026, 12:28 AM
Security Audit — agent-trust-hub — hive-query