Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
Build, audit, extend, and package GitHub-ready Model Context Protocol servers for Mailosaur. Use when Codex needs to map Mailosaur's official API or Node.js client to MCP tools, add Mailosaur email/SMS/TOTP/preview workflows, validate tool schemas against current documentation, or prepare a Mailosaur MCP repository…
Use this skill when you (an AI agent) need to orchestrate a multi-step goal by planning a DAG of tasks, executing them in dependency order, tracking state across turns, and optionally pausing for human approval — all through the dag-planner-mcp MCP tools you have available.
Foundation work for Productive MCP v2 — refactors the API client, sets up the test harness, modernizes the server bootstrap, owns transports and release plumbing. Picks up M1 (foundation fixes), M2 (McpServer migration + elicitation + completion + ResourceLinks infra), and M7 (StreamableHTTP transport + 2.0 release).
Define Model Context Protocol terms precisely (host, client, server, resource, tool, prompt, sampling, elicitation, extension, SEP). Use when the user asks what an MCP term means or how MCP concepts relate.
Operate AgentLink as a persistent, encrypted two-agent messenger. Use whenever AgentLink MCP tools are available, a peer agent is connected, or the task could benefit from asking, answering, or coordinating with a trusted peer agent through AgentLink.
Coordinate material work with humans, Codex, Claude Code, and other active coding-agent sessions through one local room per Git project. Use it to announce bounded work, inspect active handles and worktree targets, report collisions or blockers, and post provenance-rich handoffs without treating chat as authority.
Manage MCP servers with mcp-hub — discover/search servers, add and configure them from their docs, wire secrets to the OS keychain, reload, verify, and troubleshoot. Use when adding/removing/configuring an MCP server in mcp-hub, when the user says "add mcp to mcp hub", or asks how to find or manage MCP servers.
Use when coordinating with other agents or humans through the AgentTalks channel - sending messages, asking open questions, claiming shared resources, or checking who is active before touching shared state.
Use when starting to build a new MCP server, agent, skill, or prompt asset — captures the foundational asset record during the build via the coderoot-authoring MCP.
Hand a task to Claude Code for a gentle, thorough second pass. Use when you'd like a collaborative second agent to investigate, review, or implement something in the current repo — for example "ask Claude to look at this", "get a second opinion from Claude", or "have Claude take a pass at this". Claude runs in the…
Route Markdown documents, evidence-backed project facts, conversation history, personal memory, and reusable procedures without double-writing. Use when an Agent must search or write AFS documents, propose or retrieve governed facts, inspect evidence and Timeline, or choose the correct durable store.
Orchestrate local ZCode, Kimi, Claude Code, Codex CLI, and Pi agents through Agent Gateway MCP. Use for low-cost routing, multi-agent comparison, failure escalation, visible terminal demos, safe worktree-based parallelism, or maintaining the gateway.
Talk to another Claude chat through a shared room — on this machine or a different one — and pick up a conversation a previous chat left behind. Use when the person wants two chats to exchange messages, wants this chat to continue where another one stopped, mentions a room id — six words joined by dashes, like…
Crea una "neurona" (plugin) para el Motor Agéntico siguiendo el estándar de la Zona de Plugins — o un artefacto standalone para quien no use el conector. Úsalo cuando alguien quiera añadir una pieza al Motor (un panel, una herramienta, un tema, una fuente de datos) sin tocar el núcleo ni saber cómo funciona por dentro.
Build and iterate on MCP apps with custom UI views (user-generated apps) using the local MCP App Dev Kit. Covers the full local loop: brainstorm against UX guidelines, scaffold tools + views in the dev kit, persist data in a local Durable Object (DO SQL), run the dev server, expose it with ngrok, connect it to Claude…
Build MCP servers for LLM-external service integration. Use for FastMCP (Python), MCP SDK (Node/TypeScript), tool design, API integration, resource providers.
★not rated 0 1mo agoA41 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: