OCT. 15, 2026 (英语).
在今天的Google Cloud上建设的开发者可以将一个长视距内存agent从终端上安装，将零售的垂直技能安装到一条线的编码助手中，并将全wired生产服务部署到Cloud Run,Agent Runtime,Agent Garden，或Gemini Entertainment app上——都来自同一曲折源树。 这棵树是google/adk-recipes，是google/adk-samples的重组后继者，现在服务于Agent平台的一个单一的真理来源。
ADK的食谱是agent友好的：我们已经使你的agents很容易：1)发现各种agent使用案例的相关食谱，2)从你最喜欢的吊带中自动安装和使用。
google/adk-recipes将其内容分为三个顶级文件夹：核心/用于Agents CLI和DevRel团队整理的参考模式，插件/用于安装的垂直技能包，作为符合Agent插件v1.0.0 spec的插件Recipes，以及用于五种语言的端到端的可部署服务。 每个作品的设计都以机器可读性为首要考虑。
下面是走过寄存器的形状，三个行之有效的例子(每层一个)，以及对新结构所促成的反思。
三个顶级根基，每个根基基于一个读者动作：
每个食谱都有一个有效的清单 yaml, AGENTS.md，它告诉编码助理打开它是什么食谱和如何使用它，以及(用于缩写/和插件/)一个Dockerfile加一个可运行性测试。 CI在每个PR上执行标准，并在repo中执行一套编码辅助技能(prepare-python-recipe，生成-manifest,github-pr-review，等等)来进行大部分的准备，因此你的agent可以脚手架，验证，并提交食谱端到端。
每张AGENTS.md，每张明信片.yaml，每张SKILL.md主要是为AI编码助理撰写的，该助理将代表您打开重播；README.md文件是人类仍然阅读的内容。 手册的术语表捕捉到双重框架："Recipe——一个可运行的agent实例(或可导入的agent模块)在核心/, contrib/，或插件/，被ADK开发者消耗，并同样编码agents".
adk-recipes在消耗ADK的三台Agent平台表面充当ADK内容的单一真理来源。 每个都通过自己的协议读取存储器。
Agents CLI 读取核心/. 当一个开发者运行uvx——从google-agents-cli agents-cli创建我的-agent时，Agents CLI可以按要求脚手架一个核心/食谱。 今年早些时候在Google Cloud Next的CLI，发布，暴露了全脚手架部署评价出版的生命周期；它服务的基本内容来自我们的食谱。 地平线(Horizon)，我们下面审查的参考带，是一个实例图案Agents CLI作为生产模板处理。
Agent花园的特色是精心挑选。 进入寄存器的开发者-facing路径是 contrib/的牵引请求：GitHub中的一种食谱获得，成为公共ADK生态系统的一部分，从此可以被其他开发者和通过读取寄存器的编码助理发现。 Agent花园画廊的外观是由Agent花园团队单独作出的编辑决定，它整理了一组配方，并在自己的cadence上向公共目录宣传。 agent在画廊内后，Google Cloud项目的任何开发者都会得到一击安装。
在这些主表面之外，Google Cloud本身的基础设施组件在食谱本身内作为构件出现。 Horizon使用Agent平台沙盒进行每个用户的沙盒，内存库进行耐久内存，Agent平台会话进行会话状态。 零售产品搜索插件Recipe在Gemini Enterprise Agent平台上使用矢量搜索2.0与BigQuery对齐。 零售虚拟Tryon插件Recipe使用Gemini图像模型用于静态复合材料，Veo 3.1用于动画输出的参考到视频。 食谱不仅与Agent平台共存；它们记录在使用中。
对于在Google Cloud工具外(Claude Code,Antigravity,Cursor,Windsurf,Zed等)与编码助理合作的开发者，插件食谱通过vercel-labs维护的开放npx技能CLI额外安装。
核心/
构建能处理单一对话的agent是一个非常清楚的问题。 建立自己的互动日，记住用户昨天所做的以及规划明天的任务，是一项重大不同的工作。 这种长视距系统的硬部分是相同的，不管agent正在建造做什么：跨会场记忆，每个用户对工具执行的沙箱，秘密管理，防止逃跑行为的护栏，以及使多步任务可以恢复的管弦原始。 地平线(Horizon)，它生活在核心/python/长视距-harness，是我们为收集这些原始生物在一个地方建造的参照带。
从新克隆到运行中的地平线实例的最短路径：
uvx google-agents- cli 设置
gcloud auth 应用程序默认登录
gcloud 配置集工程您 PROJECT ID
gcloud服务启用aiplatform.googleapis.com 互联网档案馆的存檔，存档日期2013-03-02.
git克隆 - 深度 1 - filter=blob: none - sparse https://github.com/google/adk-recipes.git
cd 缩写
git 稀疏检查集核心/平顶/长视距
cd 核心/平顶/长视孔
# 默认：LHA ENVIRONMENT BACKEND=本地，内存会话，不需要部署云
使本地化
纯文本
复制
部署后端通过Terraform 运行；将它倒下。 Horizon的评价套件生活在测试/eval/evalsets/作为十八个命名的行为主题，包括内存召回，注射式的过滤尝试，在对抗性输入下停止的护栏，工具选择的纪律，以及旅行级的情节艺术评价。 使用uv运行的 pytest 测试/ eval 执行标语。
第一个针对运行后端的聊天回合确认内存层为wired正确：
地平线就是我们给编码助理们的绳索在建造更坚硬的东西之前可以学习当一个开发者要求他们的助手建造必须穿越数天交互的agent时，正确的下一步是uvx google-agents-cli以Horizon为脚手架创建，其余的牵引装置就位。
地平线编码了一个参考模式。 它没有编码垂直。 为此，寄存器的第二层持有插件Repipes，每个插件都涉及特定的行业领域。 这个发布中第一个完全覆盖的垂直是零售，在插件/零售/下提供。
插件Recipe格式坐落在Agent插件v1.0.0规格之上，发布于agent-plugins.org，是跨编码助理的容器安装AI技能的开放标准。 规格将插件定义为由一个或多个技能组成的捆绑，具有顶级插件。 json 显示声明名称，版本，许可证，以及所有权。 每个技能都是一个自成一体的文件夹，有一个SKILL.md对话安装器，一个README.md，一个EVAL. 质量门，脚本和资产技能需要做它的工作。 零售插件的布局是：
插件/ 零售/ 设备
QQ 插件。 json
* 技能/技能
* 产品搜索/
QQ SKILL.md #Q-MODE对话安装器
校对：Soup
* EVAL.yaml# LLM-即审即决
│ script 脚本/├──
│ ├── references/
* 资产/
虚拟行/
└── ...
今天这个插件里有两个技能飞船第一，零售-产品搜索，在Gemini企业Agent平台上搭建一个ADK agent，并配有由矢量搜索2.0支持的矢量搜索检索工具以及一个大查询目录。 该技能处理整个管道：将一个CSV或JSON目录摄入BigQuery，用Vertex AI Embeddings生成多模式和文本嵌入，构建矢量搜索集，用检索 docs工具脚手架ADK agent，并部署到Cloud Run. agent是故意的单-agent：一个root agent有一个检索工具，没有协调层。 语义复杂性坐落于检索管道和评价器面额中，而不是agent到agent的通信中；增加管线层会模糊图案而不是澄清图案。
第二个技能是零售-虚拟-Tryon，生成一个模型穿戴或持有产品的摄影现实主义复合体，上面有可选的八秒工作室猫行录影带。 它默认使用gemini-2.5-flash-image，当编辑质量重于耐久时，回落到gemini-2.5-pro-image. 对于视频，它使用veo-3.1-generate-001在参考-到视频模式中，在图像管道产生的静态上调节生成的运动。 技术船是飞行前的产品切割分类器和可配置的安全等级，在第三次实际生产部署后，操作细节变得很明显。
两种技能都安装到任何读取有单一命令的技能的编码助理中：
从反重力，Claude Code,Codex,Cursor...
npx技能添加 google/adk-recipes - 技能零售-产品搜索 - 全深度
npx技能增加google/adk-recipes - 技能零售-虚拟-trion - 充分深度
——全深旗指示CLI通过默认的顶级浏览，并发现嵌入插件/零售/技能/结构。 在~/.agents/技能/<skill-name>/中安装获得，并与本地机上每个检测到的编码助手进行对称连接：
$ npx 技能添加 google/adk-recipes - 技能零售-产品搜索 - 充分深度 - g
探测到的claude-code 2-1-288 agent Agent——非交互安装
资料来源：https://github.com/google/adk-recipes.git。
复制仓库
找到11个技能
安装到：反重力CLI，扩容，IBM Bob,Claude Code,OpenClaw
CodeBuddy 继续粉碎机器人鹅朱尼 iFlow CLI
Kilo Code, Kiro CLI, Kode, MCPJam, Mux, OpenHands, Quder, Qwen 代码
罗码，崔，Windsurf,Zencoder,......
安装摘要
~~.agents/技能/零售产品搜索 ~~
QQ普：反重力CLI,Droid,Kilo代码，Amp +15 QQ
│ symlink → Augment, IBM Bob, Claude Code, OpenClaw, CodeBuddy +25 │
├────────────────────────────────────────────────────────────────────────╯
安装了 1 个技能
完毕！使用前审查技能；它们运行时有完整的agent权限。
技能生态系统的设计是为了作曲。 零售-产品搜索 SKILL.md 指示主机编码助理在google/技能尚未安装的情况下，从这些技能中抽取两个伴机技能，将大查询基础和Gemini API使用等交叉关切保留在一个地方，并由多个垂直插件重新使用。
安装后，编码助理读取SKILL.md，并启动Q-MODE采访。 对于零售-产品搜索，快速启动是两个问题(GCP项目ID和产品数据路径，两者都有合理的默认); Full Setup 新增两个又涵盖产品领域级别和GCP区域。 到采访结束时，技能已经写了。/design-spec.md 反映项目的实际形状和已执行的脚本/设置.py 在大Query登陆目录，生成嵌入，构建矢量搜索集，并运行EVAL. 与结果相反如果标语和断言检查低于80%的通过目标，技能就会停止并暴露具体的失败。
在捆绑的五种产品演示目录上进行端到端的设置运行：
$.venv/bin/python 脚本/设置.py -- config./design-spec.md
INFO: main: SETUP PIPELLINE(由设计-spec.md驱动)
INFO：主要：项目：测试项目-425813
INFO: main：区域：我们中央1
INFO: main：验证样本数据。 验证了5个记录。
INFO: main：输入大查询。..
INFO: ingest bigquery：创建数据集测试项目-425813. retail skill 产品
INFO：摄入量最大比大：装入5个有效产品
INFO: main：创建矢量搜索收藏。..
INFO：摄入量顶点搜索：收藏成功创建
INFO：摄入量顶点搜索：摄入完成：5 创建，0 跳过，0 错误
INFO: main: SETUP (中文(简体) ).
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
OCT. 15, 2026
A developer building on Google Cloud today can scaffold a long-horizon memory agent from the terminal, install a retail vertical skill into a coding assistant with one line, and deploy a fully-wired production service to Cloud Run, Agent Runtime, Agent Garden, or Gemini Enterprise app — all from the same curated source tree. That tree is google/adk-recipes, the restructured successor to google/adk-samples that now serves Agent Platform from a single source of truth.
ADK recipes are agent-friendly: we've made it easy for your agents to 1) discover relevant recipes for various agent use cases, and 2) install and use them automatically from your favorite harness.
google/adk-recipes splits its content into three top-level folders: core/ for reference patterns curated by the Agents CLI and DevRel teams, plugins/ for installable vertical skills packaged as Plugin Recipes conforming to the Agent Plugins v1.0.0 spec, and contrib/ for end-to-end deployable services across five languages. Every piece has been designed with machine readability as the primary consideration.
What follows is a walkthrough of the repository's shape, three worked examples (one per layer), and a reflection on what the new structure enables.
Three top-level roots, each anchored on one reader action:
Every recipe ships a validated manifest.yaml, an AGENTS.md that tells the coding assistant opening it what the recipe is and how to use it, and (for contrib/ and plugins/) a Dockerfile plus a runnability test. CI enforces the standards on every PR, and a suite of coding-assistant skills in the repo (prepare-python-recipe, generate-manifest, github-pr-review, and others) does most of the preparation, so your agent can scaffold, validate, and submit a recipe end-to-end.
Every AGENTS.md, every manifest.yaml, every SKILL.md is written primarily for the AI coding assistant that will open the repo on your behalf; README.md files are what humans still read. The handbook glossary captures the dual framing: "Recipe — a runnable agent example (or importable agent module) under core/, contrib/, or plugins/, consumed by ADK developers and coding agents alike."
adk-recipes serves as the single source of truth for ADK content across the three Agent Platform surfaces that consume it. Each reads the repository through its own protocol.
Agents CLI reads core/. When a developer runs uvx --from google-agents-cli agents-cli create my-agent, Agents CLI can scaffold on demand a core/ recipe. The CLI, launched earlier this year at Google Cloud Next, exposes the full scaffold → deploy → evaluate → publish lifecycle; the base content it serves comes from our recipes. Horizon, the reference harness we examine below, is an example pattern Agents CLI treats as a production template.
Agent Garden features a curated selection. The developer-facing path into the repository is a pull request to contrib/: a recipe lands in GitHub, becomes part of the public ADK ecosystem, and from that moment is discoverable by other developers and by coding assistants reading the repository. Appearing in the Agent Garden gallery is a separate editorial decision made by the Agent Garden team, which curates a subset of recipes and promotes them to the public catalog on their own cadence. Once an agent is in the gallery, any developer in a Google Cloud project gets a one-click install.
Beyond these primary surfaces, Google Cloud's own infrastructure components appear as building blocks inside the recipes themselves. Horizon uses Agent Platform Sandboxes for per-user sandboxing, Memory Bank for durable memory, and Agent Platform Sessions for session state. The retail product-search Plugin Recipe uses Vector Search 2.0 on Gemini Enterprise Agent Platform paired with BigQuery. The retail virtual-tryon Plugin Recipe uses Gemini image models for still composites and Veo 3.1 Reference-to-Video for animated outputs. The recipes do not just live alongside Agent Platform; they document it in use.
For developers working with coding assistants outside Google Cloud's tooling (Claude Code, Antigravity, Cursor, Windsurf, Zed, and others), Plugin Recipes additionally install through the open npx skills CLI maintained by vercel-labs.
core/
Building an agent that can handle a single conversation is a well-understood problem. Building one that holds its own across days of interaction, remembering what the user did yesterday and planning tasks for tomorrow, is a materially different undertaking. The hard parts of such a long-horizon system are the same regardless of what the agent is being built to do: cross-session memory, per-user sandboxing of tool execution, secrets management, guardrails against runaway behaviour, and the orchestration primitives that make multi-step tasks resumable. Horizon, which lives at core/python/long-horizon-harness, is the reference harness we built to collect these primitives in one place.
The shortest path from a fresh clone to a running Horizon instance:
uvx google-agents-cli setup
gcloud auth application-default login
gcloud config set project YOUR_PROJECT_ID
gcloud services enable aiplatform.googleapis.com
git clone --depth 1 --filter=blob:none --sparse https://github.com/google/adk-recipes.git
cd adk-recipes
git sparse-checkout set core/python/long-horizon-harness
cd core/python/long-horizon-harness
# Default: LHA_ENVIRONMENT_BACKEND=local, in-memory sessions, no cloud deploys needed
make dev-local
Plain text
Copied
make deploy rolls the backend to Cloud Run through Terraform; make destroy tears it back down. Horizon's evaluation suite lives in tests/eval/evalsets/ as eighteen named behavioural rubrics covering memory recall, injection-based exfiltration attempts, guardrail halting under adversarial input, tool-selection discipline, and a journey-level plot-artifact evaluation. The rubrics are executed with uv run pytest tests/eval.
A first chat turn against the running backend confirms the memory layer is wired correctly:
Horizon is the harness we give coding assistants to study before building something harder. When a developer asks their assistant to build an agent that must run across days of interaction, the right next step is uvx google-agents-cli create with Horizon as the scaffold, and the rest of the harness falls into place.
Horizon encodes a reference pattern. It does not encode a vertical. For that, the second layer of the repository holds Plugin Recipes, each addressing a specific industry domain. The first vertical fully covered in this launch is retail, available under plugins/retail/.
The Plugin Recipe format sits on top of the Agent Plugins v1.0.0 specification published at agent-plugins.org, an open standard for packaging installable AI skills across coding assistants. The specification defines a plugin as a bundle of one or more skills with a top-level plugin.json manifest declaring name, version, license, and ownership. Each skill is a self-contained folder with a SKILL.md conversational installer, a README.md, an EVAL.yaml of quality gates, and the scripts and assets the skill needs to do its work. The retail plugin's layout is:
plugins/retail/
├── plugin.json
└── skills/
├── product-search/
│ ├── SKILL.md # Q-MODE conversational installer
│ ├── README.md
│ ├── EVAL.yaml # LLM-as-judge rubrics + deterministic assertions
│ ├── scripts/
│ ├── references/
│ └── assets/
└── virtual-tryon/
└── ...
Two skills ship in this plugin today. The first, retail-product-search, builds an ADK agent with a vector-search retrieval tool backed by Vector Search 2.0 on the Gemini Enterprise Agent Platform and a BigQuery catalog. The skill handles the entire pipeline: ingesting a CSV or JSON catalog into BigQuery, generating multimodal and text embeddings with Vertex AI Embeddings, building the Vector Search collection, scaffolding the ADK agent with a retrieve_docs tool, and deploying to Cloud Run. The agent is deliberately single-agent: one root_agent with one retrieval tool and no coordinator layer. The semantic complexity sits in the retrieval pipeline and the evaluator rubric, not in agent-to-agent communication; adding orchestration layers would obscure the pattern rather than clarify it.
The second skill, retail-virtual-tryon, generates a photorealistic composite of a model wearing or holding a product, with an optional eight-second studio catwalk video on top. It uses gemini-2.5-flash-image by default, falling back to gemini-2.5-pro-image when editorial quality matters more than latency. For video it uses veo-3.1-generate-001 in Reference-to-Video mode, conditioning the generated motion on the still produced by the image pipeline. The skill ships a pre-flight product-cutout classifier and configurable safety levels, operational details that become obvious after the third real production deployment.
Both skills install into any coding assistant that reads skills with a single command:
# From Antigravity, Claude Code, Codex, Cursor …
npx skills add google/adk-recipes --skill retail-product-search --full-depth
npx skills add google/adk-recipes --skill retail-virtual-tryon --full-depth
The --full-depth flag instructs the CLI to walk past its default top-level lookup and discover the nested plugins/retail/skills/ structure. The install lands in ~/.agents/skills/<skill-name>/ and is symlinked into every detected coding assistant on the local machine:
$ npx skills add google/adk-recipes --skill retail-product-search --full-depth -g
● claude-code_2-1-288_agent Agent detected — installing non-interactively
◇ Source: https://github.com/google/adk-recipes.git
◇ Repository cloned
◇ Found 11 skills
● Installing to: Antigravity CLI, Augment, IBM Bob, Claude Code, OpenClaw,
CodeBuddy, Continue, Crush, Droid, Goose, Junie, iFlow CLI,
Kilo Code, Kiro CLI, Kode, MCPJam, Mux, OpenHands, Qoder, Qwen Code,
Roo Code, Trae, Windsurf, Zencoder, …
◇ Installation Summary ──────────────────────────────────────────────────╮
│ ~/.agents/skills/retail-product-search │
│ universal: Antigravity CLI, Droid, Kilo Code, Amp +15 │
│ symlink → Augment, IBM Bob, Claude Code, OpenClaw, CodeBuddy +25 │
├────────────────────────────────────────────────────────────────────────╯
◇ Installed 1 skill
└ Done! Review skills before use; they run with full agent permissions.
The skill ecosystem is designed to compose. The retail-product-search SKILL.md instructs the host coding assistant to pull two companion skills from google/skills if they are not already installed, keeping cross-cutting concerns like BigQuery basics and Gemini API usage in one place and reusable by multiple vertical plugins.
Once installed, the coding assistant reads SKILL.md and initiates a Q-MODE interview. For retail-product-search, Quick Start is two questions (GCP project ID and the path to product data, with sensible defaults for both); Full Setup adds two more covering product-fields level and GCP region. By the end of the interview the skill has written a ./design-spec.md reflecting the project's actual shape and executed scripts/setup.py to land the catalog in BigQuery, generate the embeddings, build the Vector Search collection, and run EVAL.yaml against the result. If the rubric and assertion checks fall below the 80 percent passing target, the skill halts and surfaces the specific failures.
An end-to-end setup run on the bundled five-product demo catalog produces:
$ .venv/bin/python scripts/setup.py --config ./design-spec.md
INFO:__main__: SETUP PIPELINE (driven by design-spec.md)
INFO:__main__: Project: testproject-425813
INFO:__main__: Region: us-central1
INFO:__main__: Validating sample data... Validated 5 records.
INFO:__main__: Ingesting into BigQuery...
INFO:ingest_bigquery: Creating dataset testproject-425813.retail_skill_products
INFO:ingest_bigquery: Loaded 5 valid products
INFO:__main__: Creating Vector Search collection...
INFO:ingest_vertex_search: Collection created successfully
INFO:ingest_vertex_search: Ingestion complete: 5 created, 0 skipped, 0 errors
INFO:__main__: SETUP
Auto-fetched and translated by AI Atlas. All rights belong to the original author.