create-subagent

You create a custom AI agent for code review by defining a unique system prompt and configuration. This Skill facilitates generating specialized subagents, such as a 'code-reviewer', that operate in isolated contexts to enforce specific quality and best practice behaviors.

An AI subagent is a specialized assistant with a tailored system prompt that runs in an isolated context. You need one to preserve main conversation flow and focus the AI on specific tasks like debugging, code review, or security vulnerability scanning.

Yes, you can configure project-specific AI agents by placing them in the `.cursor/agents/` directory. This allows you to define custom AI behaviors that are scoped directly to your current project's codebase and requirements.

Yes, user-specific AI assistants are supported by storing configurations in the `~/.cursor/agents/` directory. This setup enables specialized agent behaviors to persist globally across all your local projects rather than being limited to a single repository.

You set up a security reviewer by creating a 'security-reviewer' subagent with a specific prompt to scan for vulnerabilities. This specialized AI assistant can then be configured to run whenever your code is modified in the environment.

The main limitation of isolated context subagents is that they operate independently from the main conversation flow. While this preserves context isolation, it means the subagent cannot access the ongoing dialogue history or shared state of your primary AI assistant.