postgres mcp

PostgreSQL database management and optimization assistant. Provides complete database operation capabilities: health checks, index optimization, query plan a...

This skill is coherent: it delegates work to a postgres-mcp service and requires you to install/configure that service and provide a PostgreSQL connection string. Before installing or running: (1) verify the upstream postgres-mcp project/image (ghcr.io/crystaldba/postgres-mcp and the pip package) so you trust the code you will run; (2) be careful where you store DATABASE_URL/connection strings — placing credentials in config files or command lines can expose them via disk or process listings; (3) prefer read-only mode when you only need observation/analysis and require explicit confirmation before any write/DDL operations; (4) if you do run the Docker image or pip package, follow your org policies for third-party images/packages and run them in an isolated environment if possible; (5) scheduling periodic health checks or caching schema implies persistent changes you must set up yourself — the skill will not silently exfiltrate credentials or run arbitrary downloads as part of the skill bundle.

Type: OpenClaw Skill Name: postgres-mcp-skill Version: 1.0.0 The postgres-mcp-skill bundle is a well-structured set of instructions for an AI agent to manage PostgreSQL databases via a Model Context Protocol (MCP) server. It includes comprehensive modules for health checks, index tuning, query plan analysis, and schema management (e.g., in SKILL.md and reference/pg-execute.md). The bundle prioritizes safety by instructing the agent to confirm write/DDL operations with the user, utilize read-only modes, and employ parameterized queries to prevent SQL injection. No evidence of data exfiltration, malicious execution, or unauthorized persistence was found; the setup instructions (reference/setup-postgres-mcp.md) use standard deployment methods like Docker and Python.