菜单
Use when signing up or 가입 for current VESSL Cloud, installing or authenticating vesslctl, checking organizations and teams, or preparing a first billable GPU Workspace, Job, or 기본 세팅.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Analyze live W&B Models and Weave project data using the W&B MCP server, including runs, histories, configs, traces, evaluations, artifacts, registries, comparisons, diagnoses, and reports. Use when the user asks what happened in a W&B project, which run or model is best, why a run or trace failed, how evaluations performed, what changed between versions, or to save an evidence-backed analysis as a W&B Report.
Add W&B Models experiment tracking and guarded alerts to iterative experiment workflows, including Codex or Ralph-loop autoresearch. Use when the user asks to instrument experiments, create one W&B run per candidate, group related runs, record hypotheses and code provenance, log metrics and summaries, detect failed or stalled experiments, or validate that runs are comparable and reproducible. Use wandb-project-analyst for analysis of data already stored in W&B.
Add W&B Weave tracing to an existing LLM, agent, RAG, or tool-using application with minimal behavioral change. Use when the user asks to integrate Weave, instrument code with @weave.op, trace model or tool calls, choose trace boundaries, account for framework auto-patching, protect sensitive trace data, debug missing traces, or verify a real trace tree. Keep this skill focused on tracing; route evaluations, assets, and general W&B documentation elsewhere.
Turn a world-model, 월드모델, simulation, predictive-state, embodied-agent, or intervention-model idea into ranked falsifiable research questions and a compact experiment handoff. Use when selecting a world-model research direction, deciding whether an idea needs a world model, defining an action-conditioned or counterfactual comparison, or preparing Ralphthon work for Auto Research Track 1 or Track 2.
Use when preparing Ralphthon research specifications or 연구 명세, automated experiments, a Track 1 short paper or 논문, a Track 2 review agent or 리뷰, or an evidence-backed submission under the event deadline.
Use when signing up or 가입 for W&B Cloud, configuring a personal API key, checking authentication, or creating a first experiment-tracking run or 실험 추적 project.
| name | vessl-cloud-onboarding |
| description | Use when signing up or 가입 for current VESSL Cloud, installing or authenticating vesslctl, checking organizations and teams, or preparing a first billable GPU Workspace, Job, or 기본 세팅. |
Use only the current VESSL Cloud at https://cloud.vessl.ai and the vesslctl CLI. Separate free/read-only setup verification from billable compute, storage, and Job creation. Read the official workflow and command reference when acting. For autoresearch, also read the official VESSL repositories and defer to the pinned vessl-cloud-cookbook/autoresearch recipe.
Classify the target as signup, CLI verification, cost review, or paid tutorial. Confirm the current product URL and whether vesslctl is installed without reading credentials or ~/.config/vesslctl/config.yaml.
Do not mix the current Cloud product with legacy VESSL MLOps, its vessl Python CLI, or its Organization/Project model. Current Cloud uses Organization → Team → Workspace/Job.
Never ask for or inspect passwords, MFA codes, OAuth tokens, access-token values, cookies, browser storage, or ~/.config/vesslctl/config.yaml. The user completes credentials, OAuth/SAML, CAPTCHA, email verification, legal acceptance, age confirmation, and final signup submission.
vesslctl auth login --web; the user completes the sensitive browser step. Do not fall back to terminal password entry.vesslctl auth status
vesslctl config show
vesslctl billing show
vesslctl org list
vesslctl team list
vesslctl cluster list
vesslctl resource-spec list --usable-only
Completing these checks is a valid stopping point. Do not imply that a paid tutorial is required for setup success.
Copy assets/workspace-cost-card.md and fill it from current official UI/CLI values. Before vesslctl volume create, vesslctl workspace create, or vesslctl job create, show Organization/Team, cluster/region, resource spec and GPU count, availability, hourly price, current credit, container image, storage capacity/rate/estimated duration and cost, mounts, exposed ports, expected duration, timeout behavior, and cleanup plan. Obtain explicit confirmation for that exact resource.
After approval, create the smallest suitable resource, run python -c 'print("Hello, VESSL Cloud!")', verify logs/output, and return immediately to cleanup.
For auto-research, use vessl-ai/vessl-cloud-cookbook/autoresearch pinned at 97a0af14b0acae042162b1f70f17fbe2d570afa2 as the execution source of truth. Its batch-job/prep.sh and batch-job/submit.sh create billable Jobs and mount an object volume. Use a participant-owned writable fork and an exact live single-A100 resource spec from vesslctl resource-spec list --usable-only -o json. Export that slug explicitly; never accept the recipe's H100 default or substitute another GPU.
Keep the official benchmark and batch-job scripts unchanged. The Ralphthon overlay bounds the recipe to one baseline, at most three sequential candidates, and one confirmation; it does not use async fan-out or an unbounded loop. The W&B recorder runs locally after vesslctl job logs is fetched, so no W&B API key is placed in VESSL.
Pause stops Workspace compute but can preserve state and leave storage charges. Terminate is destructive and stops the Workspace or Job; object volumes and other separately managed storage may remain. Show the consequence and obtain explicit confirmation immediately before either action unless the exact timeout termination was already part of the approved creation card. Never rely on credit exhaustion as automatic cleanup.
AUTORESEARCH_TIMEOUT_S stops only the cookbook's local polling. The remote Job keeps running and billing unless it has independently reached a terminal state. On timeout, inspect vesslctl job show <slug> -o json, show the live state, execute only the approved vesslctl job terminate <slug> action, and recheck terminal state and storage before claiming cleanup.
| Level | Evidence |
|---|---|
| Account ready | User confirms verified signup and Organization |
| CLI ready | Auth/config/org/team checks succeed without revealing tokens |
| Cost reviewed | Live cost card is complete; no compute created |
| Tutorial complete | Approved resource prints Hello, VESSL Cloud and reaches a verified cleanup state |
Return the completion level, verified evidence, waiting-for-user action, exact planned or completed side effect, hourly exposure, current billing/credit state, and cleanup state.
vessl configure → stop; that is the legacy product.resource-spec list --usable-only and prices.