Skip to content
Get packs

Get Available Resources

Lets an agent check what CPU, memory, disk and GPU it can really use before heavy local work

Source Download skill
Repository
K-Dense-AI/scientific-agent-skills
Stars
55,780
Licence
MIT
Files
8 files

What it does

When you ask for resource-aware planning or are about to run a heavy local job, the agent can run bundled Python scripts that read this machine's CPU, memory, disk, scheduler (Slurm) and container limits, plus any GPU or accelerator (CUDA, ROCm, Apple Metal) that is visible. It produces a redacted JSON summary and a conservative plan for how many workers to run, without stress-testing the machine or assuming that visible hardware is actually usable.

Great for

  • Researchers or data scientists about to run parallel jobs on a shared server, cluster allocation or container
  • People who want to avoid out-of-memory crashes or oversubscribing CPUs before a large local workload
  • Developers building resource-aware pipelines who want a standard snapshot format

Skip it if

  • Benchmarking or stress-testing hardware: the skill explicitly does not run benchmarks, and its numbers are conservative estimates, not performance measurements
  • Confirming a GPU actually works for a framework: GPU runtime compatibility is marked not_tested, so you still need to check your framework can use the device
  • Environments where sharing hostnames, paths or device IDs with the agent is not acceptable: the output is redacted but the detector still reads system information

Try asking

  • “Check how many CPU workers I can safely run for this 100-task job with 2 GB per worker”
  • “Before I start this analysis, what memory and disk do I actually have inside this container?”
  • “Am I limited by Slurm or cgroups on this node, and how many cores can I use?”
  • “Create a resource snapshot and plan a CUDA-backed workload without running any GPU tests”
Step-by-step guide: how to use Get Available ResourcesInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 8 original files, the MIT licence and an attribution note.

    get-available-resources.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip get-available-resources.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip get-available-resources.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Ask for a stdout snapshot first; only write a file if you need one, since the file is restricted to a single .json name in the current directory
  • Use --skip-accelerators if GPU management tool probes are slow, and remember GPU usability still needs a framework-level test
  • Treat plan outputs as conservative ceilings and verify memory-heavy jobs with your own small trial run

What's inside

  • references/resource_semantics.md
  • references/snapshot_schema.md
  • references/sources.md
  • scripts/_common.py
  • scripts/accelerator_diagnostics.py
  • scripts/plan_workload.py
  • scripts/snapshot_tools.py
  • SKILL.md

Redistributed unmodified from K-Dense-AI/scientific-agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.