Glossary
The AI coding vocabulary, defined.
Twelve terms you will meet within an hour of picking up Cursor, Claude Code or Codex — each defined in a paragraph you can quote, then explained tool by tool. Free to read, no account.
- Agent Skills
- Agent Skills are an open standard for extending AI coding agents with reusable, shareable capabilities. They install with
npx skills add <owner/repo>from the Skills.sh marketplace and work across Claude Code, Cursor, Codex and more than 35 other agents. They are lighter than MCP servers: a skill augments one capability. - Read the definition →
- AGENTS.md
- AGENTS.md is a project-level configuration file recognised by Codex, and increasingly by other tools, that defines how AI agents should behave in a repository. It is placed at the project root or in subdirectories, and specifies coding standards, test commands, deployment patterns and tool preferences. Its purpose matches CLAUDE.md; the convention differs.
- Read the definition →
- CLAUDE.md
- CLAUDE.md is a markdown file Claude Code reads automatically for project-specific context. It sits at the project root, or in subdirectories for scoped rules, and defines coding standards, common commands, architecture notes and any instruction the agent should always follow. It is created with
/initand updated with/memoryor the#shortcut. - Read the definition →
- Context window
- A context window is the maximum amount of text, measured in tokens, that a model can process in one request. Current Claude Fable 5, Opus 5 and Sonnet 5 tiers support 1M tokens; GPT-5.6 Sol, Terra and Luna expose 1.05M through the API; Gemini 3.1 Pro supports 1M. Host products can expose smaller effective windows than the model allows.
- Read the definition →
- Git worktrees
- Git worktrees are a Git feature that lets you check out several branches at once, each in its own directory. Codex can use worktrees for isolated tasks in the ChatGPT desktop app, which prevents conflicts when parallel local tasks run; its CLI and Cloud use different execution models.
- Read the definition →
- Headless mode
- Headless mode is running an AI agent non-interactively, typically in a CI/CD pipeline or an automation script. Claude Code supports it with
claude -p "task"; Codex supports it throughcodex execand its Cloud API. There is no human in the loop: the agent executes and exits. - Read the definition →
- Hooks
- Hooks are automation triggers that fire at specific points in an agent’s workflow. In Claude Code they run before or after particular tool calls — file edits, command execution — and are defined in
.claude/settings.json. They are used for linting, formatting, validation and custom logic the agent cannot skip. - Read the definition →
- MCP (Model Context Protocol)
- MCP, the Model Context Protocol, is an open standard created by Anthropic for connecting AI assistants to external tools and data sources. MCP servers expose capabilities — database queries, API calls, browser automation, file operations — that any MCP-compatible agent can use. Cursor, Claude Code and Codex all support it, so one server serves all three.
- Read the definition →
- Plan mode
- Plan mode is an explicit phase in which an AI coding agent breaks a complex task into structured steps before implementing anything. In Claude Code it is reached with
Shift+Tabor/plan, with/effortadjusting how hard the model thinks. Cursor and Codex expose planning through their own agent interfaces. - Read the definition →
- Sandbox
- A sandbox is an isolated execution environment that constrains what AI-generated commands can affect. In Codex the sandbox policy is surface-specific and kept separate from approval, so what may run and what must be confirmed are two decisions. Claude Code and Cursor provide their own isolation controls.
- Read the definition →
- Slash commands
- Slash commands are actions prefixed with
/that trigger built-in or custom behaviour in an AI coding agent. Claude Code includes/help,/clear,/compact,/model,/effortand/simplify; Codex includes/model,/new,/reviewand/status. Typing/shows the live list, which is the only reliable inventory. - Read the definition →
- Sub-agent
- A sub-agent is a separate AI instance spawned by a main agent to handle a specific subtask. In Claude Code sub-agents run with their own context while the parent continues; in Cursor they can be specialised for testing, documentation or refactoring, each in its own context window; in Codex parallel cloud tasks work similarly.
- Read the definition →
These twelve are the terms that come up first. The documentation carries the full A-Z — 71 entries, from Agent Mode to YOLO Mode, in English and Polish.