gpu

gpu is a command for Claude Code from Generous-Corp/pulp. It costs 20 tokens per session (355 once invoked), scanned A, original, MIT.

A command set for checking GPU health and collecting measured evidence. A GPU is a processor used for graphics and other parallel work; the commands report whether a check passed, failed, or could not be verified.

In plain words
What is it for?
Use it to list and run predefined GPU checks, inspect numeric results, preserve evidence, and capture a related trace when needed.
Why use it?
It avoids guessing from screenshots or treating unavailable hardware as tested. It also keeps the results and artifact records needed to review a diagnosis.

Command for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/generous-corp/pulp/gpu
Clone the repo
git clone --depth 1 https://github.com/Generous-Corp/pulp

Made for: Claude Code.

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.

agentmods badge for gpu

README.md
[![agentmods](https://agentmods.dev/badge/commands/generous-corp/pulp/gpu.svg)](https://agentmods.dev/commands/generous-corp/pulp/gpu)
Your own site
<a href="https://agentmods.dev/commands/generous-corp/pulp/gpu"><img src="https://agentmods.dev/badge/commands/generous-corp/pulp/gpu.svg" alt="Measured on agentmods" height="20"></a>
Per session 20 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 355 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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.

ModelPer sessionOnce invoked
Fable 5 $0.00020 $0.00355
Opus 5 $0.00010 $0.00178
Sonnet 5 $0.00004 $0.00071
Haiku 4.5 $0.00002 $0.00036

Measured today against content hash 20b01465844c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

gpu scanned grade A with 0 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 today.

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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/commands/gpu.md · 38 lines

What it actually says

Start with the cheapest truthful path:

pulp doctor gpu --json
pulp gpu recipes list --json
pulp gpu recipes list --symptom <exact-symptom-token> --json
pulp gpu recipes show <recipe-id> --json

Then run the selected closed recipe into a new path-confined artifact directory:

pulp gpu probe --recipe <recipe-id> --artifacts /tmp/pulp-gpu-evidence --json

Exit 0 is a measured pass, exit 1 is a completed measured failure, and exit 2 is unavailable or unverified. Runtime, internal-validation, and artifact publication failures also use typed unverified JSON and exit 2; never read them as completed measurements. Preserve the typed JSON, gpu_evidence_id, adapter status/class, numeric oracle, and artifact digests. An unknown or unverified adapter is not an authentic hardware claim, even when the backend is Dawn/WebGPU. Never infer correctness from a screenshot alone.

Use --negative-control only to prove that the declared mutation fails for its intended causal reason. Do not run GPU readback, validation, or artifact writing on an audio thread. For a correlated trace, capture through the canonical exact-instance trace lifecycle and use /trace with gpu-probe over the same flushed .pftrace.

The recipe catalog is Pulp's product/tool workflow. It does not authorize a generic rendering API or a DPR policy; those remain evidence-gated Vellum follow-up work.

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.

  1. today First seen · 38 lines · 20 tokens per session scan A 20b01465844c

Subscribe to this mod's changes

gpu is a command published in the GitHub repository Generous-Corp/pulp (16 stars, last pushed today), licensed MIT. It adds 20 tokens to every session and 355 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-04.