Catalog
A short list on purpose: 7 rules, 12 skills, 10 prompts, 6 recipes. Copy the file, or compose a zip that drops into .cursor.
Loaded on demand when the task matches the description.
A repeatable review: generic heroes, interchangeable cards, brochure copy, and weak empty states. Use it before you call a UI done.
A finishing workflow after the feature works. The agent should not invent new scope; it should make the existing UI feel considered.
How to name, describe, and structure a skill so Cursor picks it up for the right jobs and ignores it otherwise.
Structure, copy, and layout for a marketing page that states a job to be done. Built for Next.js and Tailwind.
A workflow for writing tasks, repos, and acceptance checks for Cursor cloud agents.
A small procedure for staging, splitting, and describing work so git history stays usable — including when Cursor made the edits.
Turn a SKILL.md into a Custom Mode so it stays in context for the session. Use when a playbook should last longer than one slash invoke.
Decide when a Cursor cloud agent should subscribe to events versus run once. Use for babysitting PRs, scheduled ingest, or waiting on Slack.
A defensive pass for a pull request: injection, broken auth, secrets, SSRF, unsafe deserialization. Complements Cursor’s Security Review bot; this file is what the agent in the editor should follow.
How to brief and run a Cursor Project: the coordinator stays on planning, shared context, and review. Use for a feature, a migration, or recurring work that should outlive one chat.
A short decision skill for where a cloud agent should execute. Cursor-hosted stays the default. Self-hosted is for code, builds, and secrets that must stay on your network.
How to brief a long Cursor session when Grok 4.7 is the model: hold the goal, verify your own work, do not turn the chat into x.ai news. Use for long coding or knowledge work inside Cursor — not for ChatGPT or Claude kits.