Cursor is my agent IDE — the place agents read the repo, run commands, and land PR-ready diffs under human review.
In my daily workflow
- I open with a clear goal, relevant @-files, and an active skill when applicable.
- I let the agent explore before editing; I reject shotgun changes.
- I run lint, typecheck, and targeted tests before accepting large diffs.
- I commit in logical chunks with messages that explain intent.
How it makes me work smarter
The IDE agent excels at cross-file refactors — gallery pages, admin tables, SEO metadata — that would take me hours manually. The constraint is review: I treat agent output like a junior PR author. Fast iteration, strict verification.
My setup
- Cursor with project rules in AGENTS.md
- MCP for external systems; terminal for builds
- Skills for repeatable workflows
- Browser MCP for visual verification on UI work
On this portfolio
This gallery — types, detail component, thirty content entries, CSS, and route wiring — is the kind of vertical slice I ship through the agent IDE: scaffold, implement, style, then verify with `tsc` and a local dev server.
