agentleFS
Sign inStart free
NVIDIAKnown publisherNVIDIA/cloudai/.agents/skills/cloudai-remote-install/SKILL.md

cloudai-remote-install

Skill103 starsChanged 58 days agoApache-2.0View on GitHub

Install or update a CloudAI checkout on a remote cluster for Slurm or standalone execution. Does not run benchmarks.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 7 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use remote home for Python projects and environments, not the shared artifact directory.

    line 9
  2. Distinctive:

    Use the first 16 hex characters of SHA-256 of local hostname + NUL + resolved checkout path, not the branch name.

    line 13
  3. Very specific:

    Use scripts/deploy.py for all remote commands and transfers, not direct SSH.

    line 18
  4. Distinctive:

    Do not update an installation while running or queued work still uses it.

    line 25
  5. Distinctive:

    Verify CLI startup and the selected system config after installation, without running benchmarks.

    line 27
  6. Distinctive:

    If none exists, inspect the cluster and adapt the closest repository example: cluster name, scheduler, partitions, installation/results …

    line 33
  7. Distinctive:

    Use NFS/Lustre for these paths.

    line 38
  8. Distinctive:

    Look for existing config and storage hints in the user's SSH configuration and cluster setup.

    line 39
  9. Very specific:

    Ask before removing each candidate, after checking running and queued jobs.

    line 53
  10. Very specific:

    Never remove ~/cloudai, active or queued-job deployments, or shared configs and artifacts.

    line 54
  11. Very specific:

    If cleanup is declined, place .cloudai-keep in that deployment and exclude it from future cleanup suggestions.

    line 56

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Installation

  • One local checkout -> one remote installation with its own uv environment. Use remote home for Python projects and environments, not the shared artifact directory.
  • Deploy the main local checkout to ~/cloudai. Never delete it or deploy a linked worktree over it. Keep it maintained without overwriting remote edits.
  • Deploy linked worktrees to ~/cloudai-worktrees/<checkout-id>. Use the first 16 hex characters of SHA-256 of local hostname + NUL + resolved checkout path, not the branch name. Reuse that directory on subsequent deployments.
  • Transfer only files needed for installation and the task, including local changes. Exclude local environments, caches, bytecode, Git metadata and local results/install artifacts. Git-based deployment is also fine.
  • Use scripts/deploy.py for all remote commands and transfers, not direct SSH. It only provides transport; perform the checks and setup described here. Invoke with Python as deploy.py HOST [--dry-run] ACTION ...: run 'COMMAND', copy '~/cloudai-worktrees/<checkout-id>/' SOURCE..., or fetch 'SOURCE' DEST. Downloads have no upload exclusions.
  • Reuse remote uv, or install it if missing. Use it to install CloudAI and manage the deployment's own environment; do not reuse another checkout's virtualenv.
  • Do not update an installation while running or queued work still uses it.
  • Preserve remote edits and personal files, including custom TOMLs. Do not prune remote-only files.
  • Verify CLI startup and the selected system config after installation, without running benchmarks.

Cluster configuration

  • Reuse the user's existing system config across installations on the same cluster. Keep shared configs outside deployment directories.
  • If none exists, inspect the cluster and adapt the closest repository example: cluster name, scheduler, partitions, installation/results paths and any required account or other cluster options. Use the current models for valid fields.
  • Different execution backends may need different configs, but configs on the same cluster should share artifact installation and results paths. Use NFS/Lustre for these paths.
  • Look for existing config and storage hints in the user's SSH configuration and cluster setup. If information is not readily available, propose what you can establish and ask the user rather than guessing or searching broadly.
  • Persist CLOUDAI_SYSTEM_CONFIG in the appropriate remote shell startup file. Feature-specific configs should be explicit overrides, not replacements for the user's default. Pass the selected config to noninteractive commands too; they may not load the shell startup file.

Cleanup

  • During remote work, check for deployments unused for more than 21 days. Touch .cloudai-last-used after installation and whenever using the deployment. Missing markers or unrecorded manual use are not proof of inactivity.
  • Ask before removing each candidate, after checking running and queued jobs. Never remove ~/cloudai, active or queued-job deployments, or shared configs and artifacts.
  • If cleanup is declined, place .cloudai-keep in that deployment and exclude it from future cleanup suggestions.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free