agentsclimarketplace

Unity vscode cursor

Skill AmberLee2427/unity-skills/skills/unity-vscode-cursor

A specialized Agent Skill designed for researchers and students using the Ohio State University (OSU) Arts and Sciences (ASC), Unity HPC cluster

Install
npx -y skills add AmberLee2427/unity-skills --skill unity-vscode-cursor

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Unity remote editor workflows for VS Code and Cursor, including acceptable usage, login-node caveats, and safer patterns.

SKILL.md

3.2 KB, as published. Nobody here has run it

Unity VS Code / Cursor

Use this skill when the user is asking about remote editing on Unity with VS Code or Cursor, especially over SSH, with a remote editor server, or from a browser/desktop session.

What Is Acceptable

  • Use remote editing for code navigation, small edits, and normal development work.
  • Use VS Code Remote SSH or Cursor remote workflows to connect to a compute node, desktop session, or other approved non-login host.
  • Use the Unity jump host only as the documented gateway into the cluster, not as a workstation or compute host.
  • Edit files on a login node only if the work is lightweight and does not start background servers, indexing, or builds.

What Not To Do On Login Nodes

  • Do not run heavy builds, tests, package installs, or model loads on login nodes.
  • Do not leave VS Code Server, Cursor servers, language servers, watchers, or large indexing jobs running on login nodes.
  • Do not treat a login node as a long-lived remote workstation.
  • Do not use the jump host for interactive development beyond the documented SSH entry point.

Safer Patterns

  • Prefer local editing plus sync when the task is mostly file editing.
  • Prefer a compute-node allocation for remote editor sessions that need indexing, language servers, or background tooling.
  • Prefer Unity OnDemand desktop sessions when you need a browser-backed GUI environment.
  • If you need notebook or app-style work, use the Unity web UI rather than forcing it through an editor server.

Remote SSH Caveats

  • Remote editor servers are real processes with CPU, RAM, and disk impact.
  • A remote server that feels local on a laptop can still be too much for a login node.
  • If the editor wants to keep a daemon alive, move that session to a compute node or desktop allocation.
  • Watch for hidden background work: file watchers, TypeScript/Python servers, search indexing, extension scans, and repo-wide symbol indexing.

Unity-Specific Guidance

  • Unity docs and team norms favor OnDemand as the easiest access path for browser-based desktop, Jupyter, MATLAB, and similar GUI workflows.
  • Unity login nodes are for light coordination, copying, and editing, not for computation.
  • Remote SSH access goes through the documented jump host and VPN/2FA path, not direct workstation-style use.
  • If the user needs a GUI, browser app, or persistent interactive environment, steer them toward Unity Desktop / Simple Desktop / Jupyter instead of trying to make VS Code or Cursor do everything.

Decision Rule

  • If the task is simple editing, remote SSH may be fine.
  • If the task needs background compute, indexing, or long-lived processes, use a compute node or OnDemand session.
  • If the task needs a browser GUI or notebook, use the Unity web UI flow.

When To Hand Off

  • Use unity for general cluster rules and routing.
  • Use unity-python-envs for package, conda, wheel, and environment setup.
  • Use unity-ondemand for web UI, desktops, and app-launch workflows.
  • Use unity-jupyter for notebook-specific launch settings and troubleshooting.
  • Use unity-slurm for batch scheduling and GPU request details.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.