The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill executes an MCP server directly from a remote GitHub repository using the uvx tool within slow_query_analyzer.py. This code originates from the official Volcano Engine (ByteDance) repository and is used to facilitate database communication.
[EXTERNAL_DOWNLOADS]: The skill downloads the necessary database interface components from github.com/volcengine/mcp-server at runtime. This is part of the standard installation and execution flow for this vendor's tools.
[COMMAND_EXECUTION]: The script invokes the uvx command-line utility to provision and run the ByteHouse MCP server as a subprocess.
[PROMPT_INJECTION]: The skill processes untrusted SQL query data from database logs, which represents a surface for indirect prompt injection.
Ingestion points: Data is read from the system.query_log table in slow_query_analyzer.py via the run_select_query tool.
Boundary markers: No specific boundary markers are used to encapsulate the query text in the output.
Capability inventory: The skill can execute SQL queries and write JSON files to the local output/ directory.
Sanitization: No sanitization or filtering of the retrieved SQL query text is performed before it is included in the analysis reports.

byted-bytehouse-slow-query