Advanced Features

Agents

Agents are autonomous assistants that can explore code, run commands, and apply edits with less back-and-forth.

When to Use

• Large tasks spanning multiple files
• Repetitive workflows (review + fix + test)
• Deep investigations where the AI needs to gather broad context first

How to Use Effectively

• Assign one clear objective per agent
• Ask for explicit deliverables (changed files, test results, risks)
• Use parallel agents for independent workstreams, then merge outcomes

Example Prompt

Use one agent to inspect API auth flow and another to inspect middleware redirects.
Each should return:
1) likely root causes
2) exact files to change
3) validation steps

Rules (Project-Specific Guidance)

Rules define persistent guidance for how AI should behave in your project (style, architecture, review criteria, and constraints).

Good Rule Categories

• Coding standards (naming, file structure, comments)
• Framework conventions (routing, state management, data access)
• Safety constraints (no secret leaks, migration policies)
• Review expectations (tests required, performance checks)

Why This Improves Productivity

• Reduces repeated prompt instructions
• Makes outputs more consistent across contributors
• Prevents common mistakes before they happen

Example Rule Ideas

• "Always add or update tests for behavior changes."
• "Prefer existing utility patterns in @lib before creating new helpers."
• "Do not introduce breaking API changes without migration notes."

Skills

Skills are reusable capability bundles (instructions, workflows, and domain-specific best practices) for repeatable tasks.

Use Skills For

• Documentation writing standards
• Release note generation
• Migration playbooks
• Security review checklists

How to Use Effectively

• Create focused skills for high-frequency tasks
• Keep examples in each skill so outputs stay predictable
• Version and refine skills as your team standards evolve

Example Prompt

Use our handbook-writing skill to produce a new API guide section.
Follow tone, structure, and examples from existing docs pages.

MCP (Model Context Protocol)

MCP connects Cursor to external tools and data sources so the AI can work with systems beyond local files.

Typical Use Cases

• Issue/PR workflows in GitHub
• Ticket lookup in project management tools
• Internal documentation retrieval
• Database or API diagnostics

Productivity Benefits

• Fewer context switches between editor and browser
• More accurate answers using live system context
• Faster incident handling with direct tool access

Effective Usage Tips

• Grant least-privilege access for safety
• Keep server descriptions clear so capabilities are discoverable
• Start read-only where possible, then enable write actions when needed

Example Prompt

Using GitHub MCP, summarize open PR comments blocking release.
Group by repository area and suggest a merge order with risks.