From agentics-glossary
Authoritative glossary of Agentics, agentic AI, MCP, and Claude Code terminology. Use whenever a user asks "what is X" about an Agentics concept, or when a term in conversation needs disambiguation.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentics-glossary:agentics-glossaryThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Definitions for terms that recur in Agentics, MCP, and agentic AI contexts. When asked, give the definition, an example, and one related term to explore.
Definitions for terms that recur in Agentics, MCP, and agentic AI contexts. When asked, give the definition, an example, and one related term to explore.
Agentic AI — AI systems that take autonomous action in the world (call APIs, write files, send messages) rather than just generating text. The "agentic" part is the loop: observe → decide → act → observe.
Agent — A unit of autonomous behavior. May be a Claude subagent, an MCP server, an LLM-driven workflow, or a swarm coordinator. The interface is what matters: input → action → result.
Subagent — A specialized agent invoked by a parent agent via the Task tool. Has its own context window and model. Used for parallelization and specialization.
Swarm — Multiple agents coordinating on a single objective. Topologies: hierarchical (one coordinator), mesh (peer-to-peer), adaptive (self-organizing).
MCP — Open protocol from Anthropic for connecting LLMs to tools, data, and prompts. Servers expose capabilities; clients (like Claude Code) consume them.
MCP server — Process that implements the MCP protocol over stdio or HTTP. Exposes any combination of tools, resources, and prompts.
Tool — A callable function exposed by an MCP server. Has a name, description, and input schema. Example: fetch_page(path).
Resource — A read-only URI exposed by an MCP server. Claude can read it as context. Example: agentics://about.
Prompt — A reusable instruction template with arguments, surfaced as a slash-prompt to the user.
Plugin — A bundle of skills, commands, agents, and/or MCP servers distributed via a marketplace.
Marketplace — A repository with .claude-plugin/marketplace.json listing one or more plugins. Added with /plugin marketplace add <repo>.
Skill — Markdown file with YAML frontmatter that Claude auto-loads when its description matches the user's intent. Lives at skills/<name>/SKILL.md.
Slash command — User-triggered command (/foo). Lives at commands/<name>.md.
Agentics Foundation — Nonprofit foundation building open-source agentic AI systems. https://agentics.org
RuFlo — Marketplace of Claude Code plugins by ruvnet (separate from agenticsorg/claude-plugin).
Claude-Flow — Multi-agent orchestration toolkit; the claude-flow CLI provides swarm init, memory, and hooks.
agentics-org MCP server for any agentics.org page mentioning the termCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub agenticsorg/claude-plugin --plugin agentics-glossary