Kipt

Your prompts and skills, one keystroke away.

Explore

google-agents-cli-scaffold skill: create and upgrade agent projects

The agents-cli scaffold skill from google/agents-cli shows coding agents how to create, enhance and upgrade agent projects, with templates, flags and targets.

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-scaffold skill?

google-agents-cli-scaffold is an Agent Skill that teaches a coding agent to set up agent projects with the agents-cli command. It covers three commands, scaffold create, scaffold enhance and scaffold upgrade, plus template choices, deployment targets and the "prototype first" workflow.

It is one of seven skills in google/agents-cli, Google's "CLI and skills for building agents on Gemini platform." The SKILL.md description says not to use it for writing agent code or for deployment operations, which have their own skills.

As of 26 October 2026, its skills.sh page listed about 511,800 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 starts with a mandatory step: before scaffolding a new project, load the workflow skill and finish its Phase 0, which clarifies what the agent should do, which tools it needs and whether you want a prototype or full deployment.

Map choices to flags

Choice Flag
Prototype, no deployment --prototype
Deployment target --deployment-target agent_runtime, cloud_run, gke or none
CI/CD runner --cicd-runner github_actions or google_cloud_build
Session storage --session-type in_memory, cloud_sql or agent_platform_sessions
A2A protocol No flag; built into the scaffolded app
RAG, sandboxed execution, memory, OAuth, guardrails No flag; these come from reference recipes

The skill notes that --datastore, the agentic_rag template, agents-cli infra datastore and agents-cli data-ingestion no longer exist. If the agent reaches for one, it wants a recipe instead.

Templates

adk (Python, the default) and adk_go (Go) are the only built-in templates. Both deploy to Agent Runtime, Cloud Run or GKE and have A2A built in. Other frameworks ship as template repos, for example --agent google/agents-cli/extensions/langchain/template@v1.11.0. Live and voice agents have no template: you scaffold adk and convert it.

Create, enhance, upgrade

  • Create takes --agent, --deployment-target, --region and usually --prototype. The skill tells the agent to pass --agent-guidance-filename to match the coding agent: GEMINI.md for Antigravity CLI, CLAUDE.md for Claude Code, AGENTS.md for Codex and others.
  • Enhance adds a deployment target or CI/CD runner to an existing project. Pass --agent-directory if the code is not in app/.
  • Upgrade moves a project to a newer agents-cli version with a smart merge, with --dry-run, --auto-approve and a --force fallback.

The CLI runs in "strict programmatic mode" by default: every required parameter has to come as a flag, or it raises a UsageError.

Scaffold as reference

When you only need a Dockerfile, Terraform or CI/CD files, the skill has the agent create a throwaway project in /tmp/, copy what it needs and delete the rest. A references/flags.md file lists every flag for create and enhance, including --bq-analytics, --skip-checks and the --adk quickstart shortcut.

How do you install it?

The Getting Started guide lists Python 3.11+, uv and Node.js as prerequisites. The recommended install adds the CLI and all seven skills:

uvx google-agents-cli setup

Or install skills only, either the whole set or just this one (the command shown on skills.sh):

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

The skill's metadata lists the agents-cli binary as a requirement, installed with uv tool install google-agents-cli. It relies on the workflow skill for Phase 0, so the full set is the practical choice. In Claude Code, run /skills to confirm the agents-cli skills are listed. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

How do you use it?

The skill loads when you ask to "create an agent project", "start a new ADK project", "build me a new agent", "add CI/CD to my project", "add deployment" or "upgrade my project", per its description. You can also call /google-agents-cli-scaffold in Claude Code.

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

  • Create a Python ADK prototype called invoice-helper. Ask me what it should do first.
  • Add Cloud Run deployment to this prototype, with Cloud SQL sessions.
  • Add a CI/CD pipeline to this project. Ask me which runner before you change anything.
  • Upgrade this project to the current agents-cli version. Show me the dry run first.
  • I only need the Dockerfile and Terraform that agents-cli would generate for GKE. Don't touch my project.

The skill's own example for a Go agent is agents-cli scaffold create my-go-agent --agent adk_go --deployment-target cloud_run --prototype. Google's Project Structure page shows the layout the scaffold produces.

Tips and limits

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

  • Don't create the folder. "NEVER mkdir before create." A pre-existing directory triggers enhance mode.
  • Ask before Git and CI/CD. The skill says never to create a Git repo or push without asking, and to present GitHub Actions and Cloud Build instead of picking one silently.
  • Agent Runtime manages sessions. If you deploy to agent_runtime, remove any session_type setting from your code.
  • Never hand-write A2A. The skill says the A2A surface is "non-trivial and changes across versions," so scaffold it instead.
  • Edit only your files. For Python, customize app/agent.py and app/tools.py; for Go, app/agent.go. .env is yours, and the rest of the template wires up serving, sessions and A2A.
  • Upgrade leftovers. --force never removes dependencies or files the new template dropped, so remove unused ones by hand.
  • Security audits. On skills.sh, Gen Agent Trust Hub rated it Pass with risk level Safe (22 October 2026), noting that its commands run locally and its downloads are standard dependencies. Socket (20 October 2026) and Snyk (risk level Low, 20 October 2026) rated it Pass.

Under Apache 2.0 you can copy and modify the skill 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 flags reference, or paste the repo link and pick from the list of all seven agents-cli skills. 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-scaffold skill?

It is one of seven Agent Skills in Google's google/agents-cli repo. It teaches a coding agent to create a new agent project with agents-cli scaffold create, add deployment or CI/CD to an existing one with scaffold enhance, and move a project to a newer agents-cli version with scaffold upgrade.

Which templates does agents-cli scaffold support?

The skill says adk (Python, the default) and adk_go (Go) are the only built-in templates, both with the A2A protocol built in. Other frameworks ship as template repos you pass to --agent, such as the first-party LangChain template in the agents-cli repo.

What is the prototype-first pattern?

The skill recommends starting with --prototype, which skips CI/CD and Terraform so you can get the agent working first. When you are ready, run agents-cli scaffold enhance . --deployment-target agent_runtime (or cloud_run or gke) to add deployment.

What are the project name rules for agents-cli scaffold create?

The skill says the name must be 26 characters or less and use only lowercase letters, numbers and hyphens. It also says not to create the directory first, because the CLI creates it and a pre-existing folder causes enhance mode instead of create mode.

How do I install the agents-cli scaffold skill?

Run uvx google-agents-cli setup to install the agents-cli command and all seven skills, or add only this one with npx skills add https://github.com/google/agents-cli --skill google-agents-cli-scaffold. The prerequisites are Python 3.11 or later, uv and Node.js.

Does scaffold upgrade overwrite my agent code?

The skill says upgrade smart-merges updates while preserving your customizations. The --force fallback overwrites scaffolding files but keeps agent code, config files, README.md, the agent guidance file and eval datasets and config. Run it with --dry-run first to see which files change.

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.