# Cursor Reference > Cursor AI 编辑器官方文档参考技能。When working with Cursor features, hooks, skills, MCP, CLI commands, rules, permissions, configuration, or any other Cursor internals — either when users ask OR when the Agent needs to know how something works. Fetch from https://cursor.com/docs as the authoritative source instead of relying on training data. Use when this capability is needed. - Skill: `tomevault-io/cursor-reference` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/cursor-reference` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/cursor-reference/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/cursor-reference --- # Cursor AI Reference Skill ## Core Principle **Always prefer fetching from official documentation over relying on training data.** When working with Cursor features, use the official docs at `https://cursor.com/docs` as the authoritative source. This applies both when users ask AND when the Agent needs accurate information about how Cursor works internally. ## When to Use This Skill **Trigger in all these contexts:** - User asks about Cursor features, configuration, or commands - Agent needs to implement hooks, skills, MCP servers, or plugins for Cursor - Agent needs to configure Cursor settings (hooks.json, mcp.json, rules) - Agent encounters an unfamiliar Cursor feature or command - Agent needs to write Cursor hooks (preToolUse, postToolUse) - Agent needs to understand permissions, allow/deny lists - Agent needs to use Cursor CLI (agent command with subcommands) - Agent needs to integrate with MCP servers - Agent needs to troubleshoot Cursor issues - Agent needs accurate API reference, SDK usage, or configuration syntax ## Official Documentation Index Base URL: `https://cursor.com/docs` ### Getting Started - Overview: `/docs` - Quickstart: `/docs/intro` - Hooks: `/docs/hooks` - Third-party hooks: `/docs/reference/third-party-hooks` ### CLI Reference - CLI overview: `/docs/cli/reference` - Parameters: `/docs/cli/reference/parameters` - Permissions: `/docs/cli/reference/permissions` - MCP: `/docs/cli/mcp` ### CLI Commands - `agent` - Main agent mode command - `agent login` / `agent logout` - Authentication - `agent status` / `agent whoami` - Status check - `agent models` - List available models - `agent mcp list` - List MCP servers - `agent mcp add` - Add MCP server - `agent mcp login ` - Login to MCP - `agent about` - About cursor - `agent update` - Update cursor - `agent ls` - List sessions - `agent resume ` - Resume a session - `agent create-chat` - Create new chat - `agent generate-rule` / `agent rule` - Generate customization rules - `agent install-shell-integration` - Install shell integration - `agent -p --force` - Headless mode for CI ### Hooks System - Hooks overview: `/docs/hooks` - Third-party hooks: `/docs/reference/third-party-hooks` - `preToolUse` hook - Called before tool execution (allow, deny, modify input) - `postToolUse` hook - Called after successful tool execution (audit, inject context) - Matcher support - Filter hooks by tool type (e.g., "Shell") ### MCP (Model Context Protocol) - MCP overview: `/docs/cli/mcp` - `agent mcp list` - List configured MCP servers - `agent mcp add` - Add new MCP server - `agent mcp login ` - MCP authentication - Configuration via `mcp.json` - Can disable MCP tools globally ### Permissions - Permissions docs: `/docs/cli/reference/permissions` - Configure allow/deny lists for: - Shell commands - File paths (glob patterns supported) - Web domains - MCP tools ### Rules - Generate custom rules with `agent generate-rule` - Customization rules for behavior ## Workflow ### Step 1: Identify the Topic Map the question to the most relevant documentation page(s). Use the index above for guidance. ### Step 2: Fetch from Official Docs Use WebFetch to retrieve the relevant page(s): ``` WebFetch(url="https://cursor.com/docs/[topic]", prompt="Extract all relevant information about [specific topic]") ``` For unknown topics, start with the main docs page: ``` WebFetch(url="https://cursor.com/docs", prompt="Find documentation related to [topic]") ``` ### Step 3: Synthesize and Answer Combine information from official docs with your own knowledge. When citing specific features, configuration options, or command syntax, **always reference the official documentation** as the source. ### Step 4: Provide Actionable Guidance Don't just quote docs — provide concrete examples, configuration snippets, or step-by-step instructions based on what the official docs say. ## Examples **User asks about hooks:** → Fetch `/docs/hooks`, explain hook types (preToolUse, postToolUse), configuration, and provide examples. **Agent needs to configure MCP:** → Fetch `/docs/cli/mcp`, explain MCP server setup, CLI commands (mcp list, mcp add, mcp login), and provide examples. **User asks about permissions:** → Fetch `/docs/cli/reference/permissions`, explain allow/deny configuration for shell, files, web, and MCP tools. **User asks about CLI commands:** → Fetch `/docs/cli/reference/parameters`, list relevant commands with flags and examples. **User asks about headless/CI mode:** → Explain `agent -p --force` for non-interactive/CI pipelines. ## Important Notes - Cursor has a hooks system similar to other AI coding tools but with differences — always fetch the specific Cursor documentation. - The CLI uses `agent` as the main command (not `cursor`), with subcommands for different operations. - Permissions use glob patterns for file paths and support granular control per category. - MCP servers are configured in `mcp.json` but managed via CLI commands. - If a page doesn't exist or can't be fetched, fall back to your training knowledge but clearly note when the information may be outdated. --- > Source: [platootalp/claude-harness](https://github.com/platootalp/claude-harness) — distributed by [TomeVault](https://tomevault.io).