An operations guide for Grid'5000, a French research testbed with clusters of physical computers reserved for computer-science experiments. It covers scheduling, remote access, distributed runs, monitoring, result syncing, and cleanup.

Use it to reserve machines, run experiments over SSH, launch distributed jobs, monitor them, and bring results back.

It explains the operational steps and usage rules needed to run experiments safely on shared research machines.

g5k — Skill for Claude Code by 0nyr

Install

Getting it into your agent

It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.

Clone the repo

git clone --depth 1 https://github.com/0nyr/g5k-kickstart

agentmods

npx agentmods add skills/0nyr/g5k-kickstart/g5k

Made for: Claude Code.

Or install g5k-kickstart, the plugin that ships this one along with the rest of its 1 skill.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/0nyr/g5k-kickstart/g5k/github.svg)](https://agentmods.dev/skills/0nyr/g5k-kickstart/g5k)

Your own site

<a href="https://agentmods.dev/skills/0nyr/g5k-kickstart/g5k"><img src="https://agentmods.dev/badge/skills/0nyr/g5k-kickstart/g5k/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/0nyr/g5k-kickstart/g5k"><img src="https://agentmods.dev/badge/skills/0nyr/g5k-kickstart/g5k.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 108 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 4,178 The whole file, excluding the scripts and references it only reads on demand.

Security scan C 2 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00108	$0.04178
Opus 5	$0.00054	$0.02089
Sonnet 5	$0.00022	$0.00836
Haiku 4.5	$0.00011	$0.00418

Measured 9d ago against content hash e6b0b6a81133, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade C, and why

g5k scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Downloads and executes remote codehighSupply chain

curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.

- **Python**: install `uv` once (`curl -LsSf https://astral.sh/uv/install.sh | sh`, then `source $HOME/.local/bin/env`), then `uv sync --frozen` in the project; prefer `uv run` over venv activation in batch jobs.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

- **Python**: install `uv` once (`curl -LsSf https://astral.sh/uv/install.sh | sh`, then `source $HOME/.local/bin/env`), then `uv sync --frozen` in the project; prefer `uv run` over venv activation in batch jobs.

skills/g5k/SKILL.md · 156 lines

How it starts

The opening of the file, as written. The whole thing — 156 lines — stays where its author put it; the contents beside it link to each section on GitHub.

g5k: Grid'5000 operations

Grid'5000 is a French large-scale testbed for experiment-driven research in computer science. Concretely, it is a set of sites (Nancy, Lyon, Rennes, Grenoble, ...), each hosting one or more clusters of bare-metal machines that you reserve through the OAR scheduler and use to run algorithmic experiments.

Authoritative external docs: Getting Started, Usage Policy, Hardware. This repository's docs/ folder holds the longer human-readable guides; example/ holds a runnable end-to-end example experiment.

User configuration (fill this in)

Adapt every command in this skill with the values below. If you are an agent and these are not filled in, ask the user before touching g5k.

Login: <login> (your Grid'5000 account name)
Project: <project> (the g5k project/team your account belongs to)
Primary site / cluster: <site> / <cluster> (e.g. nancy / gros; pick from Hardware based on what your experiments need)
SSH aliases: ~/.ssh/config entries so that ssh <site>.g5k works (see docs/getting-started.md)

For the MAMUT-routing collaboration, the recommended defaults are site nancy, cluster gros (see the worked example at the end).

Golden rules (non-negotiable)

Never run compute on the frontend (the machine you land on after ssh <site>.g5k, e.g. fnancy). The frontend is a shared gateway for preparation only: git, rsync, edits, environment bootstrap, light builds. All experiments run inside an OAR job. Running heavy background processes on the frontend violates the usage policy and can get the account restricted. If a reservation expires mid-run, stop and relaunch under a new job; never "finish it on the frontend". The policy also prohibits running VS Code Server, AI extensions, or background AI agents/LLMs on g5k machines. The intended setup is out of that rule's scope by construction: agents (Claude Code, Codex, ...) run on the user's own machine and drive g5k remotely over SSH, on the user's behalf, strictly as an experiment testbed. Nothing agent- or LLM-shaped ever executes on g5k itself. Keep it that way: never install or launch an agent, LLM runtime, or IDE server on a frontend or node.
Weekday jobs must not cross the 9:00 and 19:00 boundaries. Exceptions: a job submitted at or after 17:00 may cross 19:00 (the 17:00 to 19:00 portion is quota-free); a same-day submission may cross 9:00 (the after-9:00 portion counts against the day quota).
Day = small-scale, night/weekend = campaigns. Daytime quota (Mon-Fri 9:00-19:00, excluding French public holidays): the equivalent of 2 hours on all cores of the cluster, per user per day. On a 100+ node cluster that is a generous 200+ node-hours, so a few short multi-node jobs sit far below the ceiling; the binding constraints are the boundary rules and courtesy toward other users. Max job length is 14 h on weeknights (19:00 to 9:00) and 62 h on weekends (Friday 19:00 to Monday 9:00).
Reserved-but-idle time still counts as usage. Release unused reservations promptly (oardel), don't hoard nodes, and check oarstat -u $USER for zombie jobs at session start and end. There is no hard cap on the number of concurrent reservations; usage is governed by the quota and boundary rules, not a job count. Still, don't hoard.
Keep the remote home lean. Repos, venvs and dependencies live there; accumulated experiment outputs do not. Temporarily crossing GiB comfort levels during a campaign is fine as long as it does not last: sync results back, verify, then delete remote outputs (protocol below).
Self-audit when in doubt: usagepolicycheck -t (current usage), usagepolicycheck -l --sites <site> (violations), usagepolicycheck -v --start '... +0100' --end '...' (a time range).

Read the full file on GitHub · 156 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

9d ago First seen · 156 lines · 108 tokens per session scan C e6b0b6a81133

Subscribe to this mod's changes

g5k is a skill published in the GitHub repository 0nyr/g5k-kickstart (1 stars, last pushed 1mo ago), licensed MIT. It adds 108 tokens to every session and 4,178 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Claude Code DevOps and cloud Scientific ai-agents claude-code-plugin experiment-automation gnu-parallel grid5000

cran-extrachecks

Prepare R packages for CRAN submission by checking for common ad-hoc requirements not caught by devtools::check(). Use when: (1) Preparing a package for first CRAN release, (2) Preparing a package update for CRAN resubmission, (3) Reviewing a package to ensure CRAN compliance, (4) Responding to CRAN reviewer feedback.…

posit-dev/skills · 98 tokens

mirai

Help users write correct R code for async, parallel, and distributed computing using mirai. Use when users need to run R code asynchronously or in parallel, write mirai code with correct dependency passing, set up parallel workers, convert from future or parallel, use miraimap, integrate with Shiny or promises, or…

posit-dev/skills · 73 tokens

paper-opportunity-radar

Run a cumulative daily or retrospective sweep of research papers on a chosen topic, audit their claims, methods, integrity signals, and independent support, then identify overlooked but feasible project or business opportunities in a detailed source-grounded report. Use when asked to monitor papers every day, mine…

tamdogood/builder-essential-skills · 98 tokens

g5k