agent-fleet skillA
agent-fleet is agent-read markdown (skill) from yan-labs/yan-skills: 把机械化、大批量、对模型能力要求不高的子任务派给第三方模型(DeepSeek、Kimi/Moonshot,以及用户自备网关接入的 Gemini 或任何 Anthropic 兼容端点)去跑,省下 Claude 自己的 token 和上下文——用户已经用自己的 API Key 配置好 agent-fleet 时优先使用。底层是 Claude Agent SDK 驱动的一个 bypassPermissions 自主 Agent(能读写文件、跑 bash、多轮工具调用直到任务完成),编排(读写文件/跑 bash/多轮工具调用)全部在本机进行、不经过任何托管的编排基础设施;模型请求来源你自己选,可以是 DeepSeek/Kimi 官方端点、自备的 Anthropic 兼容网关,也可以是 Kollab 自己的公开 LLM 网关(`kollab-gateway`,用账号自助生成的 key,计费走自己 Space 额度)。当用户说"用便宜模型跑一下""省点 token""派给 DeepSeek/Kimi/Gemini""这个不需要用 Claude 做,批量处理一下就行""同时开几个后台任务分别用不同模.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# agent-fleet 个人本地 CLI 工具,路径 `/Users/kcsx/Project/kcsx/macmini/yan-skills/agent-fleet/`。核心用途: **把一个任务派给别的模型去跑,而不是自己动手**——尤其是那种"谁做都行、不需要 Claude 的判断力, 但数量大或很机械"的子任务(批量翻译、批量格式转换、大批量抄写、低难度调研/头脑风暴)。这样 可以省下 Claude 自己的 token,而不是每一步都用 Claude 亲自执行。 它是一个独立的个人工具仓库(`yan-skills`),和 Kollab 产品的代码库无关。它的信任边界是:本地 编排(Agent 读写文件、跑 bash、多轮工具调用)完全在这台机器上进行,不经过任何托管的编排基础设施; 但模型请求的来源你自己选——可以是 DeepSeek/Moonshot 官方端点、你自己搭的任意 Anthropic 兼容网关, 也可以是 `kollab-gateway`(Kollab 自己的公开 LLM 网关,用你自己账号自助生成的 `kollab_live_*` standalone key,费用走你自己 Space 的实时额度)。不管选哪个,接的都是**用户自己的**key,不存在 共享的托管账号体系。 ## 什么时候用 / 什么时候不用 **用**:任务是机械化的(不需要多少判断力就能做对)、批量的(同类任务重复很多次)、或对模型能力 要求不高(普通翻译、格式转换、常规调研摘要、批量文件级小改动),且目标模型已经在 `.env` 里配好 真实 key。想同时跑多个不同模型的任务(比如一个用 Gemini 写文案、一个用 DeepSeek 做调研,两个 并行互不干扰)是最典型的场景。 **不用**:任务需要深度架构判断、涉及本仓库(Kollab 或其它有专属规范的项目)需要遵守复杂工程规范 的改动、或者目标工作目录来路不明——agent-fleet 跑的是 `bypassPermissions` 全权限 Agent,对 prompt injection 没有免疫力(见下方「安全边界」),不适合处理完全不信任的目录。 ## 首次使用前置检查 ```bash cd /Users/kcsx/Project/kcsx/macmini/yan-skills/agent-fleet npm install # 首次使用需要装依赖 cp .env.example .env # 复制密钥模板(仅第一次) # 然后手动编辑 .env,填入真实的 DEEPSEEK_API_KEY / MOONSHOT_API_KEY / ... …
Read the whole file at its exact version.
How to install
mdr add yan-labs/yan-skills/agent-fleet@git:20260923.16c2bb7mdr add yan-labs/yan-skills/agent-fleet@sha256:8ad0f8ec7f38ba33Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_fopdwtj3oe22afy3)
1 badge views in 30 days
Versions
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260923.16c2bb7 latest | 2026-09-23 | 16c2bb7 | 16,221 B | A | view · diff |
| git:20260922.7a67fde | 2026-09-22 | 7a67fde | 15,959 B | A | view · diff |
| git:20260921.5fa78b4 | 2026-09-21 | 5fa78b4 | 13,666 B | A | view · diff |
| git:20260921.d04f2de | 2026-09-21 | d04f2de | 11,587 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (16221 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
yan-labs/yan-skills · 203 stars · license MIT · pushed 2026-09-23 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_fopdwtj3oe22afy3 GET https://markdownregistry.com/api/v1/resolve?ref=yan-labs/yan-skills/agent-fleet GET https://markdownregistry.com/api/v1/blob/8ad0f8ec7f38ba336feed1df4a97a3a7dcc48f3ba6c3576319488d39cc3874d8
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
More from yan-labs/yan-skills
Every file in yan-labs/yan-skills