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 agentmods add skills/volcengine/ark-cli/arkcli-understandnpx skills add volcengine/ark-cli --skill arkcli-understandgit clone --depth 1 https://github.com/volcengine/ark-cliWhat 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 | $0.00086 | $0.02932 |
| Opus 5 | $0.00043 | $0.01466 |
| Sonnet 5 | $0.00017 | $0.00586 |
| Haiku 4.5 | $0.00009 | $0.00293 |
Grade A, and why
arkcli-understand 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 2d 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 — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
arkcli +understand
CRITICAL — 开始前 MUST 先用 Read 工具读取 ../arkcli-shared/SKILL.md，其中包含认证闸门、API Key 错误恢复、配置排查与命令选择顺序。
CRITICAL — 执行 +understand 之前，MUST 先用 Read 工具读取 references/arkcli-understand.md（命令/flag/返回值/错误）与 references/sub-skills.md（12 个 sub-skill 的用途与期望输出形态）。禁止凭印象拼命令。
核心概念
+understand不是 12 套实现，而是 1 个引擎 + 一层语义：底层引擎就是+chat用的数据面 Responses API；每个 sub-skill 只是一条配方{模态, fallback 模型, 内置 system prompt}。Platform Profile 省略--model时必须使用当前 Profile 的Resources.Text.DefaultEndpoint；只有 Plan 类 Profile 才使用 recipe fallback 模型。- 命令形态：
arkcli +understand <sub-skill> --input @file [prompt]。args[0]命中注册表（12 个之一）→ 当作显式 sub-skill，其余位置参数当 prompt 叠加在内置 prompt 之上。args[0]不命中 → 整段位置参数都当 prompt，服务端按首个--input的文件模态自动路由到该模态的默认 sub-skill。
- 必须有
--input：sub-skill 是「对某个文件做理解」，没有--input会直接missing_input报错。纯 prompt 无法推导配方。 - 返回值与
+chat完全一致：arkcli 扁平 schema{id, model, content, reasoning_content, usage}，不是 Responses 原生output[].content[].text嵌套。详见references/arkcli-understand.md的「返回值」段。 - 多模态上传：image/video/doc 由 SDK
file://preprocessor 自动走 Files API；audio 是特例——内联为 base64 data URL，上限 25MB。详见 reference。 - 用户临时提供
--api-key/--base-url/ Endpoint 时，MUST 读取../arkcli-shared/references/execution-context.md。不要根据 Key 文本或 Endpoint 名称猜数据面/工作流。 --dry-run只在本地解析配方、当前 Profile 已持久化的默认 Endpoint、显式模型和输入引用，输出统一preview.v1；不读取在线 Endpoint/模型元数据、不上传文件、不调用 Responses API、不产生模型用量、不创建 response id 或持久化响应。在线才能补齐的值必须标为unresolved，不能把预览当服务端 validation。
快速决策（understand vs chat vs gen）
| 用户意图 | 走哪个 |
|---|---|
| 有明确产出形态的多模态理解任务（转写 / 翻译 / 字幕 / 框选定位 / GUI 操作 / 字段抽取 / 分章节总结 / 多说话人 / 会议纪要） | +understand（命中某个 sub-skill） |
开放式带图/视频对话、追问、推理、需要多轮接续（--store/--previous-response-id）、需要 tools（web_search/function）或 --text-format json_schema 严格 JSON |
../arkcli-chat/SKILL.md |
| 生成图片 / 视频（不是理解已有素材） | ../arkcli-gen/SKILL.md |
What ships with it
3 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.
- 2d ago First seen · 109 lines · 86 tokens per session scan A 90d75827d72b
arkcli-understand is a skill published in the GitHub repository volcengine/ark-cli (112 stars, last pushed 5d ago), licensed Apache-2.0. It adds 86 tokens to every session and 2,932 once invoked, about $0.0004 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.
Other skills, from other repositories
ui-ux-pro-max
UI/UX design intelligence for web, mobile, and desktop. This skill should be used when designing, building, reviewing, or fixing interfaces, including pages, components, design systems, accessibility, interaction, responsive layout, typography, color, charts, and stack-specific UI implementation. Searchable local…
baoyu-imagine
AI image generation with OpenAI GPT Image 2, Azure OpenAI, Google, OpenRouter, DashScope, Z.AI GLM-Image, MiniMax, Jimeng, Seedream and Replicate APIs. Supports text-to-image, reference images, aspect ratios, and batch generation from saved prompt files. Sequential by default; use batch parallel generation when the…
microinteractions
Design the small details -- triggers, rules, feedback, loops and modes -- that separate good products from great ones. Use when the user mentions "microinteraction", "button feedback", "loading state", "toggle design", "animation detail", "state transitions", "input feedback", "the interface feels dead", "make the UI…
aws-wechat-article-images
公众号封面｜公众号配图｜公众号插图｜AI 生图 — 公众号 AI 封面与配图生成，按文章标题与内容自动匹配画风，一稿多方案，多风格预设可复用。面向公众号编辑、自媒体、品牌设计。触发词：「封面」「配图」「插图」「生成图片」「给文章加图」「做个封面」「文章插图」「配个图」。不写正文只发一组图请走 aws-wechat-sticker；需要多环节串联（写+审+排+配图+发）请走 aws-wechat-article-main。.
aws-wechat-article-publish
公众号发布｜公众号草稿箱｜公众号群发｜图文推送｜微信 API｜wechat automation｜WeChat API automation｜auto publish｜scheduled publish — 公众号 API 发布工具，图文入草稿箱或直接群发，支持封面素材上传、发布前检查与 draft/published 模式切换。面向公众号运营、自动化内容团队、开发者。触发词：「发布」「提交」「群发」「推送」「发出去」「上传到公众号」「发到公众号」「可以发了吗」「发布前检查」。需要多环节串联（写+审+排+配图+发）请走 aws-wechat-article-main。.
animation-system
Use when implementing animations — AnimationPlayer, AnimationTree, blend trees, state machines, sprite animation, and code-driven animation.