Skip to content
Get packs

Agent Platform GenAI Inference

Lets an agent call Gemini and open models (Llama, DeepSeek, Qwen) on Google Cloud via the GenAI, OpenAI, or Vertex AI SDKs.

Source Download skill
Repository
google/skills
Stars
21,782
Licence
Apache-2.0
Files
9 files

What it does

Once installed, the agent knows how to connect to Google Cloud's Agent Platform, authenticate with your Google credentials, and send prompts to Gemini models or third-party open models such as Llama, DeepSeek and Qwen, then return the answers. It also provides ready-made Python examples for three SDKs and guidance for troubleshooting common errors like 429 quota or 404 not found. Before running any inference that may cost money, the skill requires the agent to show you the project, region, model, SDK and prompt and wait for your explicit approval.

Great for

  • Developers who want to test Gemini or open models on Google Cloud from an agent
  • Teams comparing model outputs across GenAI SDK, OpenAI SDK and Vertex AI SDK
  • Users who need help debugging authentication, regional endpoints or quota errors on Agent Platform

Skip it if

  • Deploying models to endpoints, which the skill explicitly excludes (use a deployment skill instead)
  • Running model evaluations, also explicitly excluded
  • Users without a Google Cloud project and billing set up, since every call is billed to that project and the scripts will fail without credentials

Try asking

  • “Ask Gemini 2.5 Pro a question in my project my-project in us-central1”
  • “Run a test prompt against DeepSeek V3.2 on Agent Platform using the OpenAI SDK”
  • “Generate code that calls a Llama model through the GenAI SDK”
  • “Why am I getting a 429 Resource Exhausted error when calling Gemini?”
Step-by-step guide: how to use Agent Platform GenAI InferenceInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 9 original files, the Apache-2.0 licence and an attribution note.

    agent-platform-inference.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip agent-platform-inference.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip agent-platform-inference.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Give the agent an explicit model ID, region and SDK, since it will ask for confirmation before running any inference and will not guess.
  • Check the Gemini and OpenMaaS model documentation links first; the skill notes its model ID examples may be outdated.
  • Confirm your project and billing before approving calls, because each inference request is charged to your Google Cloud project and the verification script runs every example.

What's inside

  • scripts/gemini_genai_sdk.py
  • scripts/gemini_openai_sdk.py
  • scripts/gemini_vertexai_sdk.py
  • scripts/openmaas_genai_sdk.py
  • scripts/openmaas_openai_sdk.py
  • scripts/openmaas_vertexai_sdk.py
  • scripts/requirements.txt
  • scripts/verify_all.sh
  • SKILL.md

Redistributed unmodified from google/skills under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.