The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the ByteHouse MCP server implementation directly from the vendor's official GitHub repository (volcengine/mcp-server) during execution. This is an intended part of the skill's functionality to enable interaction with the data warehouse.
[COMMAND_EXECUTION]: The diagnostic script utilizes uvx to launch the external MCP server tool. The command is hardcoded to use the official vendor repository, ensuring the execution environment remains within the vendor's ecosystem.
[DATA_EXFILTRATION]: The skill requires database access credentials (host, port, user, password) provided via environment variables. These credentials are used locally to authenticate the diagnostic tool with the target ByteHouse cluster and are not transmitted to unauthorized third parties.
[SAFE]: The skill processes diagnostic information from ByteHouse system tables (e.g., system.parts, system.mutations). Although this represents a surface for potential indirect prompt injection, it is considered safe in this context as the data originates from a controlled database environment and is used for reporting purposes.
Ingestion points: Diagnostic data from system.parts, system.mutations, system.replicas, and query_log tables in cluster_diagnostics.py.
Boundary markers: Not present.
Capability inventory: Executes vendor-specific MCP tool via uvx in cluster_diagnostics.py.
Sanitization: Data is formatted into JSON reports without explicit sanitization.

byted-bytehouse-diagnostics