AI Agent
AI 智能体
aka: Agent · 智能体 · AI 代理 · autonomous agent · LLM agent
An AI Agent is a software system built on a large language model that perceives context, plans multi-step actions, calls tools, and executes tasks toward a goal with a degree of autonomy. Unlike a single prompt-response exchange, an agent maintains state across steps, decides which tools or Skills to invoke, observes results, and iterates until the objective is met.
AI Agent（智能体）是构建在大语言模型之上的软件系统，能够感知上下文、规划多步行动、调用工具，并以一定自主性朝目标执行任务。与单轮提示-响应不同，Agent 在多步之间维持状态，自主决定调用哪些工具或 Skills，观察结果并持续迭代，直到达成目标。
See also: agent-skill, mcp, tool-use
Agent Skill
Agent 技能（Skill）
aka: AI Skill · Skill · 技能 · AI 技能 · Claude Skill · Copilot Skill
An Agent Skill is a structured, reusable instruction package that teaches an AI agent how to perform a specific task or domain workflow expertly. A Skill typically bundles a role definition, a step-by-step procedure, output conventions, and constraints — often as Markdown plus optional scripts and resources — so the agent behaves like a domain expert rather than a generic assistant. Skills are portable across compatible agents and composable: an agent can load many Skills and invoke the right one for the job.
Agent Skill（技能）是一种结构化、可复用的指令包，教会 AI Agent 如何专业地完成某项任务或领域工作流。一个 Skill 通常封装了角色定义、分步流程、输出规范与边界约束——常以 Markdown 加可选脚本与资源的形式存在——使 Agent 表现得像领域专家，而非通用助手。Skills 可在兼容的 Agent 之间移植，并可组合：Agent 可加载多个 Skills，并在合适的场景调用对应技能。
See also: ai-agent, skill-marketplace, mcp, prompt-engineering
Skill Marketplace
技能市场
aka: Skills marketplace · Skill registry · 技能市场 · 技能仓库 · Skill 商店
A Skill Marketplace is a curated registry where developers discover, install, share, and rate Agent Skills. It standardizes packaging and installation (e.g. a one-line install command) so a Skill authored once can be reused across teams and AI tools. skills.ifq.ai aggregates high-quality Skills across the ecosystem and adds editorial curation, categories, and trust signals such as install counts and stars.
技能市场（Skill Marketplace）是一个经过策划的注册中心，开发者可在其中发现、安装、分享并评价 Agent Skills。它标准化了打包与安装方式（例如一行安装命令），使一次编写的 Skill 可在团队与各类 AI 工具间复用。skills.ifq.ai 聚合生态中的高质量 Skills，并提供编辑精选、分类，以及安装量、Star 等可信度信号。
See also: agent-skill, ai-agent
Model Context Protocol (MCP)
模型上下文协议（MCP）
aka: MCP · Model Context Protocol · 模型上下文协议
The Model Context Protocol (MCP) is an open standard that lets AI agents connect to external tools, data sources, and services through a uniform interface. Where a Skill encodes know-how (how to do a task well), MCP provides the plumbing (how the agent reaches a database, API, or file system). Skills and MCP are complementary: a Skill can direct an agent to call MCP-exposed tools as part of its workflow.
模型上下文协议（MCP）是一项开放标准，让 AI Agent 通过统一接口连接外部工具、数据源与服务。如果说 Skill 编码的是「专业知识」（如何把任务做好），MCP 提供的则是「管道」（Agent 如何访问数据库、API 或文件系统）。两者互补：Skill 可在其工作流中指引 Agent 调用 MCP 暴露的工具。
See also: ai-agent, tool-use, agent-skill
Tool Use
工具调用
aka: function calling · tool calling · 工具使用 · 函数调用
Tool use (also called function calling) is the capability that lets an AI model invoke external functions — search, code execution, API calls — and incorporate the results into its reasoning. It is the mechanism by which an agent acts on the world. Skills frequently orchestrate tool use, specifying which tools to call, in what order, and how to interpret their output.
工具调用（也称函数调用，function calling）是让 AI 模型调用外部函数——搜索、代码执行、API 调用——并将结果纳入其推理的能力。它是 Agent 作用于外部世界的机制。Skills 常常编排工具调用，规定调用哪些工具、以何种顺序，以及如何解读其输出。
See also: ai-agent, mcp
Prompt vs. Skill
提示词与技能的区别
aka: prompt engineering · system prompt · 提示词工程 · 提示词
A prompt is a single instruction sent to a model; prompt engineering is the practice of phrasing it well. A Skill goes further: it is a versioned, reusable, structured artifact — role, workflow, output spec, and guardrails — that an agent loads on demand. Where a clever prompt lives in one conversation, a Skill is packaged knowledge that any compatible agent can install and apply consistently across tasks and teams.
提示词（Prompt）是发给模型的单条指令，提示词工程则是把它写好的实践。Skill 更进一步：它是可版本化、可复用、结构化的产物——包含角色、工作流、输出规范与护栏——供 Agent 按需加载。巧妙的提示词只存在于一次对话中，而 Skill 是被打包的知识，任何兼容 Agent 都能安装并在不同任务与团队间一致地应用。
See also: agent-skill, ai-agent