Mastra Agent Skills
mastra-ai/skillsMastra's official skills: a framework guide that routes to current docs and ten references, and a Factory supervisor skill for the mastra api CLI.
With caveats
Broad access Depth 3/5
Google agents-cli Claude skills
Google's CLI and seven skills that teach a coding agent to scaffold, build, evaluate, deploy and observe ADK agents on Google Cloud.
Our verdict Recommended with caveats
Well-built, official and evaluated, but it is a Google Cloud on-ramp: deployment needs a cloud project, and the CLI sends opt-out usage signals.
What Google agents-cli does
agents-cli is a Python command-line tool (google-agents-cli on PyPI) plus a set of skills that let a coding agent build agents with Google's Agent Development Kit (ADK) and run them on Google Cloud. The commands cover the whole lifecycle: create and enhance a project from a template, run the agent with a prompt, lint, evaluate with datasets and LLM-as-judge metrics, deploy, and publish to Gemini. The seven skills map to those phases: workflow (the development lifecycle and rules about preserving code), adk-code (ADK Python and Go API references and recipes), scaffold, eval, deploy, publish and observability. The workflow skill is meant to stay active, tells the agent to ask clarifying questions first and to load the right skill before each phase, and refuses to write agent code before a project is scaffolded. Deployment targets are Agent Runtime, Cloud Run and GKE, with Terraform and CI/CD pipelines; observability covers Cloud Trace, logging and BigQuery analytics.
Skills average about 15 KB, most with reference folders, and the repo has CI and about 30 eval files. Two extra skills adapt the setup for Claude Code and LangChain. Installation is uvx google-agents-cli setup, or npx skills add google/agents-cli for the skills only. The README says it works with Antigravity CLI, Claude Code, Codex and other agents.
Good for
- Teams building ADK agents who want to deploy on Google Cloud
- Developers who want an agent to follow a scaffold, eval, deploy workflow
- Anyone adopting Cloud Run, GKE or Agent Runtime for agents
Not for
- Projects not targeting Google Cloud or Gemini
- People who want a framework-neutral agent skill
What's inside depth 4/5
16 skills averaging about 15 KB with reference folders, a real CLI, CI and roughly 30 eval files.
- Skills
- 16
- Reference docs
- 69
- Scripts
- 0
- Tests / evals
- 37
- Templates
- 0
- Avg SKILL.md
- 15.3 KB
Counted from the repository file tree. How we score depth.
Dependencies and costs
- agents-cli (uv tool install google-agents-cli), Python 3.11+, uv and Node.js
- A Google Cloud project and credentials for deployment and cloud features (an AI Studio API key is enough for local create, run and eval)
- Telemetry: The agents-cli tool sends one usage signal per command (command name, exit code, error class, duration, CLI and Python versions, OS), written as a log entry to your own Google Cloud project's Cloud Logging, never file contents or arguments. Opt out with AGENTS_CLI_TELEMETRY=0; agents-cli info shows the current state.
Security review
Review before installingGoogle's Agents CLI skills for scaffolding, evaluating and deploying ADK agents to Google Cloud. The CLI sends usage metadata by default (opt out with an environment variable), the deploy skill requires explicit human approval before deploying but tells the agent to pass --no-confirm-project (skipping the CLI's project prompt) when the project is resolved automatically, and commands act on your Google Cloud project with your credentials. The CLI code itself is not in our copy, so only the Markdown skills were checked.
Worth knowing
- Sends usage data by default (can be turned off)The agents-cli tool sends one usage signal per command by default: a log entry (command name, exit code, error class, duration, CLI/Python versions, OS) written to your own Google Cloud project, plus a User-Agent / X-Goog-Api-Client header with versions, OS, architecture, command name and a random per-invocation ID that Google receives; opt out with AGENTS_CLI_TELEMETRY=0 or DO_NOT_TRACK=1 and check with
agents-cli info.README.md:170 - Runs remote code at an unpinned versionInstall is
uvx google-agents-cli setup(latest published version) ornpx skills add google/agents-cli; the skills also tell the agent tocurl https://adk.dev/llms.txtand WebFetch docs pages for reference.README.md:34 - Relies on a CLI or package we could not reviewAll real behavior lives in the google-agents-cli Python package, gcloud, uv and Node tooling, none of which is in our copy; the skills drive scaffold, deploy (Cloud Run, GKE, Agent Runtime) and eval commands that act on your Google Cloud project and may create billable resources.README.md:34
- Makes network requestsSkills fetch documentation from adk.dev and Google Cloud docs and call Google Cloud APIs through gcloud / the CLI with your Application Default Credentials.skills/google-agents-cli-adk-code/SKILL.md:65
What we checked
The repository (557 files) on Oct 29, 2026. Not a full review: 463 of 557 files were not downloaded (large repository; we reviewed the skill folders); it relies on code outside the repository: google-agents-cli Python package (src/google/agents/cli/ code, 463 files including .claude-plugin/plugin.json and docs hooks are missing from our copy), gcloud and the Google Cloud services it uses; No executable files were present in the files we have, so nothing was code-read; skill and reference Markdown was inspected by grep and targeted reading.
NVIDIA SkillSpector static scan: 89 pattern hits (Excessive Agency 36, Privilege Escalation 19, Tool Misuse 13, MCP Rug Pull 8, Data Exfiltration 6, Agent Snooping 3). Hits are mostly deployment-guide content: Secret Manager and CMEK references, '--skip-checks' / '--skip-auth' CLI options, curl examples against localhost or your own Cloud Run URL, and Claude-on-Cloud-Run guidance about the agent's Bash tool and secrets. The adk.dev/llms.txt fetch is a docs index, not instructions.
A review, not a certification. Skills change after we check them — read the files before installing.
License
Apache-2.0 — commercial use allowed.
Can it run unattended?
Not really. The workflow starts with a design dialogue and deployment touches real cloud resources, so it expects a human at the key steps.
Skills in this repository
Linked skills have their own review.
Alternatives to Google agents-cli
Google Agent Skills
google/skillsGoogle's official collection of about 180 skills for Google Cloud, Gemini APIs, Analytics, Data Manager, sign-in and other Google products.
Recommended
Broad access Depth 3/5
Vercel AI SDK skills
vercel/aiThe AI SDK repo's skills: one public ai-sdk skill for building with the SDK plus about a dozen maintainer-only skills for the monorepo.
With caveats
Check first Depth 2/5
Cloudflare Skills
cloudflare/skillsCloudflare's official 19-skill pack for building on Workers, Durable Objects, the Agents SDK, Cloudflare One and more, with an optional bundled MCP server.
Recommended
Check first Depth 4/5
FAQ
What is Google agents-cli?
It is a CLI and skill set from Google that helps a coding agent create, evaluate, deploy and monitor agents built with the Agent Development Kit on Google Cloud.
How do I install agents-cli?
Run uvx google-agents-cli setup to install the CLI and skills into your coding agents, or npx skills add google/agents-cli for the skills alone. You need Python 3.11+, uv and Node.js.
Does agents-cli collect telemetry?
Yes, one usage signal per command (command name, exit code, duration, versions, OS) is written to Cloud Logging in your own Google Cloud project, without file contents or arguments. Set AGENTS_CLI_TELEMETRY=0 to opt out.
agents-cli vs Mastra skills: which should I use?
agents-cli builds ADK agents and deploys them to Google Cloud. Mastra's skills target the Mastra TypeScript framework, so choose by framework and cloud rather than by skill quality.
Sources: GitHub repository (stars, license, file tree, last push Oct 27, 2026); install counts from skills.sh via Best Skills (CC BY 4.0). Review based on the README and SKILL.md files as published.