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.
npx skills add volcengine/ark-cli --skill arkcli-deploygit clone --depth 1 https://github.com/volcengine/ark-cliWrote 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.
[](https://agentmods.dev/skills/volcengine/ark-cli/arkcli-deploy)<a href="https://agentmods.dev/skills/volcengine/ark-cli/arkcli-deploy"><img src="https://agentmods.dev/badge/skills/volcengine/ark-cli/arkcli-deploy/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.
<a href="https://agentmods.dev/skills/volcengine/ark-cli/arkcli-deploy"><img src="https://agentmods.dev/badge/skills/volcengine/ark-cli/arkcli-deploy.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00274 | $0.05085 |
| Opus 5.5 | $0.00110 | $0.02034 |
| Sonnet 5.5 | $0.00055 | $0.01017 |
| Haiku 4.5 | $0.00027 | $0.00509 |
Grade A, and why
arkcli-deploy 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 17d 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.
How it starts
The opening of the file, as written. The whole thing — 140 lines — stays where its author put it; the contents beside it link to each section on GitHub.
arkcli +deploy
CRITICAL — 路由例外必须先于任何命令:用户明确要求脚本化 / CI / 无护栏 / 原始 raw CRUD 创建 Endpoint 时,立即读取 ../arkcli-infer-endpoint/SKILL.md 并由它接管;在完成交接前禁止认证检查、模型查询或其他命令。
前置: 先用 Read 读 ../arkcli-shared/SKILL.md 获取共享认证/配置/写操作守卫规则。
创建意图中的模型澄清
只要用户的最终目标仍是“创建 / 新建 / 部署 Endpoint”,即使尚未给出完整模型 ID,也必须留在 arkcli-deploy 工作流;arkcli-models 此时只是临时调用的只读候选查询能力,不能把创建任务改路由成纯模型发现。
候选必须来自本轮 ArkCLI 的实时结构化输出,禁止从模型记忆、示例或旧版本号补全。
查询预算是当前用户回合恰好一次 Bash 调用。 在执行前一次性确定 keyword、过滤条件和 --size。models search 的 keyword 是单个 catalog 子串,不要把多个概念拼成带空格的短语:先选最能缩小范围的一个 ASCII token,其余条件放到同一次调用的 flags 或 jq 本地过滤。例如“豆包代码模型”用 keyword code,再在同一管道按 name 的 doubao|seed 过滤;“生图”用 seedream。不确定稳定 token 时宁可省略 keyword 并使用已有结构化 filter,禁止把未翻译的中文短语直接提交后再换词重试。若担心输出过长,把字段投影直接放进同一条管道,例如
arkcli models search "<keyword>" --size 0 --format json | jq -c '{items: [(.items // [])[] | {name, primary_version, lifecycle_status, input_modalities, output_modalities}]}'。
这次调用无论成功、空结果、截断还是失败,都不得换关键词、调大分页、重跑同一命令或为了重新格式化再调用一次 ArkCLI;只能使用已捕获的 stdout,信息不足就如实停止。
- 完全没给模型:执行一次有界查询,例如
arkcli models search --size 10 --format json。若用户已说明用途或模态,把对应的 keyword /--modality加进同一次查询。 - 只给品牌、系列或家族名(例如 “Doubao”“Seed 2”):执行一次
arkcli models search <keyword> --size 10 --format json;家族名不是可直接传给--model的完整 ID。 - 从同一次返回的
items中读取name、primary_version、lifecycle_status和模态等已有字段。优先保留lifecycle_status=Published且名称不含internal/test的候选;状态缺失或非 Published 时只能如实标为未核实,不能称为“可部署”。完整模型 ID 按模型查询契约确定:primary_version非空时使用返回值精确拼成<name>-<primary_version>,为空时才使用<name>;这是结构化字段组合,不是从名称或日期规律猜版本。不要自行给name拼pro、lite或任何未返回的版本后缀。 - 0 个可用候选:说明本轮没有查到,并请用户补充用途、模态或关键词;不要猜一个继续。
- 1 个可用候选:复述本轮返回的完整 ID,请用户确认;若用户已要求本轮只查询,则停在这里。
- 多个可用候选:列出精简候选及完整 ID,请用户明确选择;若候选仍过多,按用户用途缩小范围,不能擅自选第一项。
澄清阶段的收敛边界:本回合第一次 search 返回后,禁止再次执行 models search,也不要对候选循环执行 models get、价格查询或其他详情调用。只有用户选定单个候选后又明确要求比较某个缺失属性,才在后续回合追加一次针对性查询。模型尚未唯一确定前,禁止执行 +deploy、infer endpoint create 或 Raw API 创建。
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 17d ago Changed · +29 lines a45f2f9bbe81
- 22d ago Changed · +2 lines 9013d416e0ce
- 1mo ago Changed df31ef0814de
- 1mo ago Changed · +22 lines · +66 tokens per session 8fdbd061e631
- 1mo ago First seen · 87 lines · 208 tokens per session scan A 963b31ec17c7
arkcli-deploy is a skill published in the GitHub repository volcengine/ark-cli (143 stars, last pushed 8d ago), licensed Apache-2.0. It adds 274 tokens to every session and 5,085 once invoked, about $0.0011 per session on Opus 5.5. 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.
Other skills, from other repositories
skypilot-multi-cloud-orchestration
Multi-cloud orchestration for ML workloads with automatic cost optimization. Use when you need to run training or batch jobs across multiple clouds, leverage spot instances with auto-recovery, or optimize GPU costs across providers.
hugging-face-model-trainer
This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV…
serving-llms-vllm
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
modal-serverless-gpu
Serverless GPU cloud platform for running ML workloads. Use when you need on-demand GPU access without infrastructure management, deploying ML models as APIs, or running batch jobs with automatic scaling.
tensorrt-llm
Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.
lambda-labs-gpu-cloud
Reserved and on-demand GPU cloud instances for ML training and inference. Use when you need dedicated GPU instances with simple SSH access, persistent filesystems, or high-performance multi-node clusters for large-scale training.