from ay-skills58
Progressive MCP client that connects to external MCP servers on-demand, lists available tools, and executes MCP tool calls without bloating context windows.
A lightweight universal client for connecting to MCP servers (Zapier, Sequential Thinking, GitHub, etc.) that loads tool schemas on-demand to avoid large context payloads. Provides commands to list configured servers, enumerate tools, and call server tools programmatically via a bundled Python script.
Use this skill when you need to interact with external MCP servers, discover available tools, or execute MCP tool actions from an agent while keeping the agent context small. Typical triggers: connecting to Zapier, listing MCP tools, or invoking a remote server action.
Works with agents that can run Python scripts and call external services (Claude Code, CLI-capable agents, Copilot-like agents).
Universal MCP client skill that connects to any MCP server (stdio, SSE, streamable HTTP, FastMCP) with progressive tool disclosure. The Python script is well-structured with clean async patterns, proper error handling, and multiple config resolution paths. Exited with usage help (no config file present in test env), which is expected behavior. No security concerns found — no hardcoded credentials, no dangerous shell patterns, no exfiltration risk.
Clean, production-quality MCP client. Progressive disclosure approach (listing tools on demand rather than dumping all schemas) is a smart design choice for context efficiency. Config resolution chain is well-thought-out with clear priority order.
