Kipt

Your prompts and skills, one keystroke away.

Explore

google-agents-cli-adk-code skill: ADK patterns for your coding agent

The ADK code skill from google/agents-cli gives coding agents an ADK cheat sheet for agents, tools, callbacks and state in Python and Go, plus install steps.

By the Kipt team · Updated 26 October 2026

Source: google/agents-cli · Maintainer: Google · License: Apache-2.0 · Checked: 26 October 2026

What is the google-agents-cli-adk-code skill?

google-agents-cli-adk-code is an Agent Skill that teaches a coding agent how to write agent code with Google's ADK (Agent Development Kit). It bundles cheat sheets for Python and Go, covering agents, tools, multi-agent orchestration, callbacks, state and the graph-based Workflow API, plus an index of worked reference recipes.

It is one of seven skills in google/agents-cli, which the README describes as "The CLI and skills for building agents on Gemini platform." The README is clear that this is a tool for coding agents, not one itself: "agents-cli is a tool for coding agents, not a coding agent itself." It names Antigravity CLI, Claude Code and Codex as supported agents, "and any other coding agent."

As of 26 October 2026, its skills.sh page listed about 513,300 installs (snapshot, 26 October 2026). The SKILL.md metadata lists Google as author and version 1.11.0.

What does it do?

The SKILL.md is short. Most of the content sits in six reference files the agent reads when it needs them. The skill tells the agent to activate /google-agents-cli-workflow first, which covers the development phases.

It does three things:

  1. Study recipes before building. For Python, it tells the agent to read the topic index in references/samples.md before answering "how do I build X." That index maps capabilities to recipes in google/adk-samples (on GitHub the repo now redirects to google/adk-recipes), such as RAG, cross-session memory, OAuth user consent, approval gates and safety plugins. For Go, it points to the upstream examples/ folder in google/adk-go.
  2. Gate code on a scaffolded project. "Do NOT write agent code until a project is scaffolded." The agent runs agents-cli info, then agents-cli scaffold create <name> --agent adk --deployment-target <target> for a new project (--agent adk_go for Go) or agents-cli scaffold enhance . for existing code.
  3. Answer from cheat sheets, then the docs. For anything beyond the cheat sheets, it fetches the ADK docs index at https://adk.dev/llms.txt or inspects the installed package source for exact signatures.

The reference files:

File Language What it covers
adk-python.md Python Core API: Agent, tools, callbacks, plugins, state, artifacts, SequentialAgent / ParallelAgent / LoopAgent, custom BaseAgent, A2A, event-driven agents, managed agents
adk-python-workflows.md Python Graph Workflow API (ADK Python 2.0): nodes, edges, fan-out and fan-in, human-in-the-loop
adk-python-live.md Python Live and voice agents: run_live, LiveRequestQueue, voice config, session limits
adk-go.md Go Core Go API: llmagent, tools, callbacks, sessions, serving over HTTP, A2A
adk-go-workflows.md Go Graph Workflow API for Go
samples.md Python Topic-indexed catalog of the ten core/python recipes in adk-samples

The samples file is strict about guessing. It says the ten recipes are "the complete set" and that names like core/python/code-execution "do not exist."

How do you install it?

The Getting Started guide lists Python 3.11+, uv and Node.js as prerequisites, with the Google Cloud SDK and Terraform as optional extras for deployment. The recommended install puts the agents-cli command and all seven skills in place:

uvx google-agents-cli setup

The guide also lists pipx install google-agents-cli && agents-cli setup and a skills-only route:

npx skills add google/agents-cli

To add only this skill, use the command shown on skills.sh:

npx skills add https://github.com/google/agents-cli --skill google-agents-cli-adk-code

The skill's metadata lists the agents-cli binary as a requirement, installed with uv tool install google-agents-cli, so the skills-only route still needs the CLI for scaffolding. The guide says macOS, Linux and Windows through WSL 2 are supported, and "Native Windows is not officially supported." For Claude Code, it says to run /skills and look for google-agents-cli-workflow and the other skills. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

For a model to run, the guide says agents-cli picks up existing gcloud Application Default Credentials, or you can set a GEMINI_API_KEY from AI Studio.

How do you use it?

The skill is model-invoked: its description lists trigger phrases such as "write agent code", "build an agent with ADK", "add a tool" and "create a callback." It also says not to use it for scaffolding or deployment, which belong to google-agents-cli-scaffold and google-agents-cli-deploy. You can call it directly with /google-agents-cli-adk-code in Claude Code.

Example prompts (suggestions, not quotes from the repo):

  • Build an ADK agent in Python that answers questions from our docs folder. Check the recipes first.
  • Add a tool to this agent that looks up an order by ID, and a before_tool callback that blocks refunds over $500.
  • Turn this single agent into a SequentialAgent: one agent drafts, a second reviews against a checklist.
  • Scaffold a Go ADK agent and add an approval step before it sends any email.
  • Which adk-samples recipe covers memory across conversations? Read its AGENTS.md and adapt the pattern here.

The skill is a reference, not a step-by-step workflow. Pair it with the workflow skill, which the docs describe as "Always active."

Tips and limits

From the SKILL.md, the repo docs and skills.sh, not from testing:

  • Model names change. The Python cheat sheet's examples use gemini-3.8-flash. Check the current Gemini models list before you pin one in production.
  • Recipes need a clone. samples.md tells the agent to sparse-clone google/adk-samples into /tmp and read each recipe's AGENTS.md. The agent needs git and network access for that step.
  • Python is the default. The recipe catalog covers Python only; for Go the skill relies on upstream examples.
  • Telemetry. The README says agents-cli sends one usage signal per command to your own Google Cloud project by default, with some metadata in request headers that Google receives. Set AGENTS_CLI_TELEMETRY=0 or DO_NOT_TRACK=1 to opt out.
  • Security audits. On skills.sh, Snyk rated it Warn (audited 20 October 2026), flagging medium-risk third-party content exposure (W011) because the workflow fetches documentation from adk.dev/llms.txt and web pages. Socket rated it Pass (20 October 2026), and Gen Agent Trust Hub rated it Pass with risk level Safe (22 October 2026).

The skill runs no scripts of its own; commands come from the agents-cli tool it calls. Under Apache 2.0 you can copy and modify it as long as you keep the license and notices and mark files you change.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link to import it with its references/ files. Paste the repo link instead and Kipt shows a list of all seven agents-cli skills to pick from. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Google.

Similar skills

Questions

What is the google-agents-cli-adk-code skill?

It is one of seven Agent Skills in Google's google/agents-cli repo. It gives a coding agent a quick reference for writing code with ADK (Agent Development Kit): agent types, tool definitions, orchestration, callbacks, state and the graph Workflow API, with cheat sheets for Python and Go and a catalog of reference recipes.

How do I install the ADK code skill?

Google's docs recommend uvx google-agents-cli setup, which installs the agents-cli command and all seven skills. For skills only, run npx skills add google/agents-cli, or add just this one with npx skills add https://github.com/google/agents-cli --skill google-agents-cli-adk-code. The prerequisites are Python 3.11 or later, uv and Node.js.

Do I need a Google Cloud account to use it?

Not for local work. The agents-cli README says create, run and eval work locally with an AI Studio API key; deployment and cloud features need Google Cloud. The skill itself is reference material, but it expects a scaffolded project and the agents-cli command before writing agent code.

Does the ADK code skill support Go?

Yes. It ships references/adk-go.md and references/adk-go-workflows.md alongside the Python cheat sheets, and it points the agent to the examples folder in the google/adk-go repo. New Go projects are scaffolded with --agent adk_go.

Why does the skill tell my agent to scaffold first?

The SKILL.md says not to write agent code until a project is scaffolded. The agent runs agents-cli info to check for an existing project, agents-cli scaffold create for a new one, or agents-cli scaffold enhance for existing code, so the code lands in the layout the rest of the toolchain expects.

More in Skill directory

All guides

Keep your prompts and skills in one place, in your own GitHub. Install every skill with one command.

Get started

Kipt is an independent product, not affiliated with or endorsed by Anthropic, OpenAI, Cursor or GitHub. Claude, ChatGPT, Codex, Cursor and GitHub are trademarks of their owners.