Cursor
Cursor
Section titled “Cursor”Cursor is an AI-native code editor built on VS Code. It supports the Agent Skills standard with both its own .cursor/ folder and .claude/ compatibility, custom slash commands, subagents with async delegation, and a Plugins marketplace. The Agents Window (v3.0, April 2026) adds parallel agent execution across repos and worktrees.
Skills
Section titled “Skills”Supports Agent Skills standard: ✅ Yes — primary .cursor/skills/, compatibility .claude/skills/
| Scope | Path |
|---|---|
| Project (primary) | .cursor/skills/ |
| Project (compatibility) | .claude/skills/ |
| Global (primary) | ~/.cursor/skills/ |
| Global (compatibility) | ~/.claude/skills/ |
Cursor looks for skills in .cursor/skills/ first, then in .claude/skills/ for compatibility — similar to GitHub Copilot. If you use .claude/skills/ as your canonical location, no symlink is needed; Cursor will find your skills automatically. Skills are loaded when they match your request, based on the description field in each SKILL.md frontmatter.
Plugins (March 2026): Cursor’s Plugins system integrates skills as first-class plugin components. The Cursor Marketplace distributes plugin packages that bundle skills alongside subagents, MCP servers, hooks, and rules into one-click installs. Skills installed via plugins follow the same SKILL.md format and discovery mechanism — the plugin system supplements the folder convention, it does not replace it. Project-scoped skills committed to .claude/skills/ continue to work unchanged alongside plugin-delivered skills.
📖 Cursor Agent Skills Documentation · Cursor Marketplace
For more about how skills work in factory engineering, see Skills.
Commands
Section titled “Commands”Folder location: .cursor/commands/ (project) or ~/.cursor/commands/ (global)
Invocation: /command-name — filename without .md becomes the slash command. Use slash-command at-artifact (e.g. /write-spec @submit-sales-order). Follow the recommended pattern (state what the user supplies; stop and prompt if missing). Do not rely on $ARGUMENTS — commands are shared via symlinks and not all IDEs support it.
Usage:
/write-spec @submit-sales-orderSince Cursor uses .cursor/commands/ and the canonical commands live in .claude/commands/, create a symlink:
ln -s ../.claude/commands .cursor/commandsBuilt-in commands (v3.0): Cursor 3 added /worktree (creates an isolated git worktree so agents work without conflicts) and /best-of-n (runs the same prompt across multiple models in parallel worktrees and compares results). These are built-in commands, not user-defined — your custom commands in .cursor/commands/ continue to work alongside them.
📖 Cursor Custom Commands Documentation
For more about how commands work in factory engineering, see Commands.
Agents
Section titled “Agents”Supports true agents: ✅ Yes
Feature name: Subagents + Plugins
Storage location: .cursor/agents/ (project); ~/.cursor/agents/ (global)
Cursor reintroduced agent support in v2.4 (January 2026) with subagents — independent agents that handle discrete parts of a parent agent’s task. Each subagent runs in its own context with configurable prompts, tool access, and model selection.
Built-in subagents: Explore (codebase navigation), Bash (shell commands), and Browser (browser interactions).
Custom subagents are Markdown files with YAML frontmatter stored in .cursor/agents/ (project-scoped) or ~/.cursor/agents/ (global). Frontmatter fields include name, description, model (inherit, fast, or a specific model ID), readonly, and is_background.
Async subagents (v2.5): Subagents can run asynchronously — the parent continues working while subagents execute in the background. Subagents can also spawn their own subagents, creating a tree of coordinated work.
Plugins (v2.5): The Cursor Marketplace packages skills, subagents, MCP servers, hooks, and rules into one-click installs.
Agents Window (v3.0, April 2026): A standalone workspace for running many agents in parallel across repos, worktrees, and cloud environments. Agent Tabs display multiple chats side-by-side in a grid. New commands include /worktree and /best-of-n. Design Mode lets you annotate UI elements in the browser to give agents precise visual feedback.
Memory: Instruct agents via .cursor/rules or AGENTS.md to read from a markdown file at the start of work and append learnings at the end.
📖 Cursor Subagents Documentation · Cursor Changelog
For more about how agents work in factory engineering, see Agents.
Workflows
Section titled “Workflows”Orchestration support: ⚠️ Partial
Cursor’s Agents Window (v3.0, April 2026) lets you run many agents in parallel across repos, worktrees, and cloud environments. A parent agent can delegate work to custom subagents defined in .cursor/agents/, and subagents can spawn their own subagents (v2.5+), creating a tree of coordinated work.
However, Cursor does not have a dedicated orchestrator that reads a workflow document and delegates to named specialists based on its contents. The delegation is ad-hoc — the parent agent decides how to split work based on its prompt, not by following a structured workflow file. You can approximate workflow-driven orchestration by writing detailed instructions in a command file and relying on the parent agent to follow them, but there is no enforcement layer.
Closest pattern: Write a command in .cursor/commands/ that describes phases and specialist subagents. Invoke it with /command-name @artifact. The parent agent reads the command and delegates to subagents, but routing logic depends on the model’s interpretation rather than a built-in orchestration engine.
📖 Cursor Subagents Documentation · Cursor Agents Window
For more about how workflows work in factory engineering, see Workflows.