golden-eval

golden-eval is a skill for Claude Code, Codex from nebius/nebius-physical-ai. It costs 46 tokens per session (1,363 once invoked), scanned A, original, Apache-2.0.

A small tested rerun that checks whether a container image actually works, usually by running its own smoke test on a real GPU. A container image is a packaged environment used to run software consistently.

In plain words
What is it for?
Use it to list or inspect image checks, run local or serverless evaluations, test batches of images, and validate the offline evaluation manifest in CI.
Why use it?
It separates problems in the image from problems in the surrounding pipeline before larger workflows are run. Its manifest also records whether each image is ready, GPU-dependent, blocked upstream, or needs an update.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit Use it to list or inspect image checks, run local or serverless evaluations, test batches of images, and validate the offline evaluation manifest in CI.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/nebius/nebius-physical-ai/golden-eval
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.

Any agent
npx skills add nebius/nebius-physical-ai --skill golden-eval
Clone the repo
git clone --depth 1 https://github.com/nebius/nebius-physical-ai

Made for: Claude Code, Codex.

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 golden-eval

README.md
[![agentmods](https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/golden-eval/github.svg)](https://agentmods.dev/skills/nebius/nebius-physical-ai/golden-eval)
Your own site
<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/golden-eval"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/golden-eval/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.

agentmods 80×15 button for golden-eval

Your own site · 80×15
<a href="https://agentmods.dev/skills/nebius/nebius-physical-ai/golden-eval"><img src="https://agentmods.dev/badge/skills/nebius/nebius-physical-ai/golden-eval.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 46 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,363 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
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 5.1 $0.00046 $0.01363
Opus 5 $0.00023 $0.00681
Sonnet 5 $0.00009 $0.00273
Haiku 4.5 $0.00005 $0.00136

Measured 11d ago against content hash d9c3e28cc82a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

golden-eval 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 11d 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.

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.

skills/tools/golden-eval/SKILL.md · 122 lines

How it starts

The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Golden eval (does this container actually work?)

A golden eval is the minimal tested rerun that proves one container image is functional. It answers a narrow, valuable question — does this image run its own smoke on a real GPU? — without standing up a cluster or a workflow. Reach for it after building or republishing an image, and when triaging whether a failure is the image or the pipeline around it.

The manifest is npa/src/npa/smoke/golden_evals.yaml (format npa_golden_evals_v1); every container in CONTAINER_IMAGE_NAMES must have an entry, enforced by npa/tests/smoke/test_golden_eval_manifest.py.

Inspect before running

npa workbench golden-eval list                 # every container, kind, gpu, status
npa workbench golden-eval list --output json
npa workbench golden-eval show <container>     # full safety + Physical AI record

Read the status column before spending anything:

  • ready — runnable now.
  • gpu-gated — needs a real GPU; a local run without one proves nothing.
  • blocked-on-upstream — excluded from batch runs unless you pass --include-blocked. A failure here is expected and is not your regression.
  • needs-image-update — the manifest and the published image disagree; rebuild before drawing conclusions.

kind tells you what is actually exercised: container-smoke, server-smoke, entrypoint-smoke, workflow-smoke, or build-import. A build-import passing means the package imports, not that the tool works. gpu is required, optional, or none.

Three execution tiers, cheapest first

npa workbench golden-eval run <container>                      # dry run (default)
npa workbench golden-eval run <container> --execute            # local runtime
npa workbench golden-eval run <container> --serverless         # one GPU, real image
npa workbench golden-eval run <container> --serverless --gpu h200 --timeout 40m

Dry run is the default and prints the command. Use it to see exactly what would execute — often enough to answer a question without running anything.

Read the full file on GitHub · 122 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.

  1. 11d ago First seen · 122 lines · 46 tokens per session scan A d9c3e28cc82a

Subscribe to this mod's changes

golden-eval is a skill published in the GitHub repository nebius/nebius-physical-ai (27 stars, last pushed today), licensed Apache-2.0. It adds 46 tokens to every session and 1,363 once invoked, about $0.0002 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 2026-08-30.