Claude Skills by nebius

Use when working on Isaac Lab RL simulation, deployment, SkyPilot workflows, or customer custom-fork support.

Use when working on LanceDB vector storage, table creation/querying, LeRobot/BDD100K imports, UDF backfills, materialized views, CLIP embeddings, or AV/perception data flows.

Use when building, launching, operating, reviewing, or live-testing the NPA LeIsaac browser teleoperation tool, its secure agent-relay transport, immutable LeRobot datasets, custom bundles, or PAIDF export/materialization.

Use when working on LeRobot workbench training, evaluation, serving, inference, dataset conversion, or robot policy workflows.

Use when deploying, launching, or reviewing the Lichtblick web viewer — an open-source, Foxglove-compatible MCAP / ROS-bag / robotics log viewer served from S3 artifacts.

Use when working on MJLab locomotion evaluation, SONIC checkpoint scoring, SkyPilot MJLab YAMLs, or Workbench MJLab CLI behavior.

Use for Nebius runtime configuration, provision-if-absent setup, cluster, registry, storage, GPU routing, and credential assumptions that affect NPA runs.

Use when operating the NPA agent VM, chat UX, API grounding, bootstrap deployment, or verify-live checks.

Use when onboarding or running the Open Dreamer (next-state/open-dreamer) JAX/Flax Dreamer 4 world-model pipeline as a multi-GPU BYOF registry candidate — causal video tokenizer, action-conditioned latent dynamics, and a real >=2 GPU data-parallel smoke.

Run or review the native Ray Train synthetic distributed CUDA reference, its S3 checkpoints, optimizer recovery, and factual Rerun metrics.

Use when working on Workbench motion retargeting, SONIC retargeted motion artifacts, SkyPilot retargeting YAMLs, or retargeting CLI behavior.

Use to run RoboCasa kitchen-task simulation as a first-class NPA workbench tool — Gymnasium task registration, kitchen asset availability, headless EGL environment reset, and random rollouts with video artifacts, through the npa-robocasa service.

Use when generating adversarial scenarios via RL, ranking mined failures of a policy-under-test, or wiring the adversarial-scenario-hardening workflow.

Use when running or debugging how the engine renders and submits SkyPilot from an npa.workflow spec: invocation, SkyPilot limits, JobGroups, multi-node tasks, runner scripts, and cleanup. For authoring, use author-npa-workflow.

Use when working on SONIC whole-body-control training, export, evaluation, serving, GPU routing, validation, or CUDA alignment.

Use to deploy or operate a Nebius soperator (Slurm-on-Kubernetes) cluster from npa — the npa.soperator/v0.0.1 spec, multi-preset worker pools, per-pool Docker/Enroot image cache, quota preflight, and post-deploy fixes.

Use for zero-GPU hosted inference through Nebius Token Factory — captioning, batch text generation, and Cosmos physical-AI reasoning — including key setup, model selection, and the npa.workflow toolRefs that need no cluster.

Use to score rollouts with a vision-language model and turn the score into a pipeline gate — single rollout, prefix-wide loop, rubric/threshold benchmark sweeps, backend selection (self-hosted, api, stub), and judging against a plan an earlier stage wrote.

Use when packaging, running, reviewing, or extending the Alibaba Wan 2.2 TI2V-5B BYOF solution, its official video artifacts, or its verified Rerun evidence.

Use when adding, changing, deploying, or calling any NPA workbench tool; captures the API/CLI/SDK/container architecture and data-flow contract.

Use when navigating, reviewing, or changing the compositional Sim2Real workflow, stateless stage adapters, durable standard-runtime resume, ComponentRecords, and S3 lineage.

Use when adding a new workbench tool to npa end to end — implementation, CLI, SDK, toolRef catalog, container, tests, docs, and skill — in the order that keeps every CI gate green.

Use when deploying, tearing down, or reproducing a fresh NPA agent VM from scratch — npa-driven destroy/fresh-setup, profile selection, tiered verify gates, and teardown failure recovery.

Use when verifying that a bootstrapped NPA agent VM can create, validate, plan, provision for, or run npa.workflow YAMLs.

Use when authoring, validating, or reviewing NPA workflow specs (apiVersion npa.workflow/v0.0.1) — declarative state machines that invoke workbench tools via SkyPilot.

Use when onboarding an OSS repo via BYOF — containerize on Ubuntu or Isaac Lab, push to an operator-controlled or authorized GHCR registry, and smoke on live Kubernetes.

Use when packaging, validating, or operating the NVIDIA Content Agents workflow that enriches a self-contained USD object with real Material Agent, Physics Agent, OVRTX, and Validation Agent stages and emits a narrow Isaac rigid-object handoff.

Use when an external contributor or maintainer adds or changes an NPA workbench image and must carry it through licensing review, trusted public-dev build, real validation, and digest-identical GHCR release promotion.

Use when running or modifying Cosmos3 inference through NPA, including Nano diffusion continuation/augmentation, framework generation, prompt/input handling, and effective guardrail or sampling arguments.

Use when writing, validating, or submitting an npa.workflow/v0.0.1 YAML that runs Cosmos 3 generation (text2image / image2image / text2video / image2video / video2video) in the npa-cosmos3 container. This is the declarative npa.workflow spec, NOT the SkyPilot YAML.

Use when planning, reviewing, or explaining Cosmos3 supervised fine-tuning and post-training in NPA, including upstream recipes, dataset/checkpoint preparation, and why NPA does not expose post-training as a fake skill command.

Use when turning an architecture diagram plus a step-by-step write-up into a working npa.workflow/v0.0.1 YAML — parse boxes/arrows/decision-diamonds and numbered steps into states, loops, gates, and catalog toolRefs, then validate/plan until green. Generalizes across sim2real, AV perception, RL, and Cosmos pipelines.

Use when authoring, reviewing, or operating an NPA workflow whose real run outputs should become a reviewable Rerun .rrd recording with factual timelines, provenance, declared artifacts, and independent content validation.

Use on a fresh machine or a new Nebius project to get from zero to a first verified result — an ordered, gated path through install, configure, credential preflight, cheapest-proof workload, then cluster provisioning, with an explicit stop condition at every step.

Use when inventing a new npa.workflow/v0.0.1 pipeline from the tool catalog — creative stage graphs, loops, gates, and reference YAML output.

Use when reconstructing real sensor captures into renderable 3D scenes with NVIDIA Omniverse NuRec / the Neural Reconstruction Engine (NRE) on Nebius — NCore V4 input, 3DGUT Gaussian training, renderable USDZ, novel-view rendering, and the Rerun recording the NPA agent displays. Also use when an NCore sequence will not load in NRE, when picking the GPU for a reconstruction, or when changing the nurec workbench tool, CLI, or SkyPilot workflow.

Use when onboarding and containerizing a world model (learned action-conditioned simulator — Dreamer/Genie/Cosmos-style latent video prediction) into NPA as a multi-GPU BYOF registry candidate. Generalizes the Open Dreamer onboarding into a reusable playbook — containerize the repo, stage a real dataset, encode the train to tokenize to dynamics to dream to visualize loop as capability smokes, and validate on real GPUs.

Use when evaluating and onboarding an open-source Physical AI solution into the NPA registry/catalog with documented capabilities, BYOF packaging, smoke tests, and live Nebius validation.

Use when authoring, running, submitting, or viewing the NVIDIA Physical AI Data Factory blueprint on Nebius + SkyPilot (no OSMO) — annotate → Cosmos Transfer augment → Cosmos Evaluator gate → re-label → Cosmos Curator + FiftyOne curate → Rerun visualize — implemented as an npa.workflow that composes existing workbench tools.

Use when designing, reviewing, or operating sim-to-real workflows that move data through simulation, policy training, synthetic generation, evaluation, and loop control.

Operate the compositional 14-stage Sim2Real npa.workflow on Kubernetes through the standard SkyPilot runtime, durable S3 resume ledger, real component images, and artifact audit.

Use when planning, reviewing, or operating robot policy training across LeRobot, Isaac Lab, SONIC, and workflow YAMLs.

Use when working on NPA reference workflow specs, runner scripts, cookbooks, customer-adaptable pipeline implementations, or the guarded examples that are not workflow authoring surfaces.