Aris Compute Guard

What it does

When your agent uses it

• : about to run experiments
• Deploy training
• Any GPU-intensive task

Example prompts

• “/aris-compute-guard”

Requirements

• Python 3
• Pre-approved tools (allowed-tools): Bash(nvidia-smi*), Bash(python*), Bash(ssh*), Bash(echo*), Bash(which*), Bash(command*), Read, Grep, Glob

Workflow steps

5 steps, taken from the step headings in SKILL.md.

1. 1Detect Target Environment
2. 2Check Compute Availability
3. 3Decision Gate
4. 4Stop and Report (when compute unavailable)
5. 5Proceed Summary (when compute available)

What it can do on your machine

Read from SKILL.md and the folder at commit d51b64e. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves these tools, so the agent can use them without asking each time:

  • Bash(nvidia-smi*)
  • Bash(python*)
  • Bash(ssh*)
  • Bash(echo*)
  • Bash(which*)
  • Bash(command*)
  • Read
  • Grep
  • Glob

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  Shell commands in SKILL.md call:

  • python3
  • ssh
  • modal

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md. Its commands use ssh, which can reach the network depending on how they are called.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Aris Compute Guard loads about 1.8k tokens when it runs. Until then it costs about 103 tokens; SKILL.md has 515 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~103

When it runs · the whole SKILL.md, loaded when a task matches

~1.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from OpenLAIR/dr-claw at commit d51b64e, republished under its MIT licence (© OpenLAIR). 515 words, ~1,835 tokens.

name

aris-compute-guard

description

Mandatory pre-flight compute resource check before running experiments. Detects whether local/remote GPU or compute resources are actually available. If resources are unavailable, STOPS the experiment pipeline immediately and reports to the user — preventing the model from hallucinating fake experiment results. Use when: about to run experiments, deploy training, or any GPU-intensive task.

allowed-tools

Bash(nvidia-smi*), Bash(python*), Bash(ssh*), Bash(echo*), Bash(which*), Bash(command*), Read, Grep, Glob

argument-hint

[environment-type]

license

MIT

metadata.author

wanshuiyin/ARIS

metadata.version

1.0.0

# Check if nvidia-smi exists
which nvidia-smi 2>/dev/null
# If exists, check GPU status
nvidia-smi --query-gpu=index,name,memory.used,memory.total,utilization.gpu --format=csv,noheader 2>/dev/null

python3 -c "
import torch
mps_available = hasattr(torch.backends, 'mps') and torch.backends.mps.is_available()
print(f'MPS_AVAILABLE={mps_available}')
if mps_available:
    print('COMPUTE_OK=true')
else:
    print('COMPUTE_OK=false')
" 2>/dev/null

# Check if any GPU framework is available
python3 -c "
import torch
cuda = torch.cuda.is_available()
mps = hasattr(torch.backends, 'mps') and torch.backends.mps.is_available()
print(f'CUDA={cuda}, MPS={mps}')
if not cuda and not mps:
    print('COMPUTE_OK=false')
    print('REASON=No GPU available (no CUDA, no MPS). CPU-only execution is not suitable for ML training experiments.')
else:
    print('COMPUTE_OK=true')
" 2>&1

# Fallback: check for nvidia-smi directly
nvidia-smi 2>/dev/null || echo "COMPUTE_OK=false"
echo "REASON=Neither nvidia-smi nor PyTorch found. Cannot verify GPU availability."

# Check SSH connectivity (timeout 10s)
ssh -o ConnectTimeout=10 -o BatchMode=yes <server> "echo CONNECTED" 2>/dev/null
# If connected, check GPU
ssh -o ConnectTimeout=10 <server> "nvidia-smi --query-gpu=index,memory.used,memory.total --format=csv,noheader" 2>/dev/null

# Check for running instances
cat vast-instances.json 2>/dev/null
# Or query Vast.ai API
vastai show instances 2>/dev/null

# Check Modal CLI is installed and authenticated
modal token verify 2>/dev/null || echo "MODAL_NOT_CONFIGURED"

⚠️ COMPUTE RESOURCES UNAVAILABLE — Experiment Stopped

I checked the compute resources and they are NOT available for running experiments.

**Environment:** [local / remote / vast.ai / modal]
**Issue:** [specific reason — e.g., "No GPU detected", "SSH connection failed", "All GPUs fully occupied"]

**What you need to do:**
- [Actionable step 1 — e.g., "Ensure your machine has a CUDA-compatible GPU"]
- [Actionable step 2 — e.g., "Free up GPU memory by stopping other processes"]
- [Actionable step 3 — e.g., "Configure a remote server in CLAUDE.md"]

**Alternative options:**
- Set `gpu: modal` in CLAUDE.md to use Modal serverless GPU (no local GPU needed)
- Set `gpu: vast` in CLAUDE.md to rent an on-demand GPU from Vast.ai
- Configure a remote GPU server with `gpu: remote` in CLAUDE.md

I will NOT proceed with running experiments or generating results, as doing so without actual compute resources would produce fabricated output. Please resolve the compute issue and try again.

✅ Compute resources verified:
- Environment: [local / remote / vast.ai / modal]
- GPU: [GPU name, count, free memory]
- Status: Ready for experiments

Proceeding with experiment execution.

/aris-compute-guard
/aris-compute-guard local
/aris-compute-guard remote

© OpenLAIR, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/aris-compute-guard of OpenLAIR/dr-claw.

Open the folder on GitHubat commit d51b64e