Claude Code Glossary
Plain definitions for the vocabulary. What each term means, and how it fits the way a Flow fires the right skill.
FLOW.md
A FLOW.md is a routing file that maps a task or intent to the skill that should handle it. Flowy reads it on each prompt and auto-invokes the right skill, so you do not have to remember which one to call.
Auto-invocation
Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.
Claude Code hook
A Claude Code hook is a script that runs on a lifecycle event, such as every user prompt or after context compaction. Flowy uses a hook to re-assert its routing, which keeps the right skill firing across a long session.
SKILL.md
A SKILL.md is a single reusable instruction set that teaches Claude Code how to do one job well, such as writing a failing test first. You install it and it can be invoked by name, or auto-invoked by a FLOW.md.
MCP (Model Context Protocol)
MCP, the Model Context Protocol, is an open standard for connecting AI agents to external tools and data sources through a consistent interface. An MCP server exposes capabilities that an agent like Claude Code can call.
Context compaction
Context compaction is when a long conversation is summarized to fit the model’s context window. It frees space but can drop detail, which is why instructions read once at the start tend to lose force after a compaction.
Claude Code plugin
A Claude Code plugin is a packaged bundle that installs one or more components into Claude Code: skills, agents, commands, hooks, or MCP servers. Installing a plugin adds all of its components at once.
Claude Code skill
A Claude Code skill is a reusable instruction set (a SKILL.md) that teaches the agent how to perform a specific job. Skills are the building blocks a Flow hand-picks and routes so the right one fires for each task.
