Agent Platform Tuning Management skill (Google): list, check and cancel tuning jobs

Google's skill that makes Claude list, inspect and cancel GenAI tuning jobs in Agent Platform with the Python SDK.

Installs: 8.9K; on skills.sh
Stars: 21.8K; on GitHub
Updated: Nov 2026; last repo commit

agent-platform-tuning-management makes Claude manage GenAI tuning jobs in Agent Platform with the Python SDK. It can list jobs, show one job's state and base model, and cancel a job that is running too long. It does not start tuning: SKILL.md sends that to agent-platform-tuning.

Listing and getting run without asking. Cancelling needs a preview card and your typed confirmation in a separate turn. It needs a Google Cloud sign-in, a Project ID, a region and the google-cloud-aiplatform package.

Good for

Google Cloud users who run GenAI model tuning jobs
Anyone who needs to check the state of a tuning run
Teams that want a job cancelled only after a typed confirmation

Install agent-platform-tuning-management

Claude Code

Copy this prompt.

Install the agent-platform-tuning-management skill in Claude Code. Run this command:

npx skills add https://github.com/google/skills --skill agent-platform-tuning-management

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

The skill runs Python and gcloud on your machine, so there is no Claude app route. Update installed skills later with npx skills update. Before first use, sign in with gcloud auth login and gcloud auth application-default login (SKILL.md Phase 0); if import vertexai fails, the skill runs pip install google-cloud-aiplatform. You need a Project ID and a region.

From the vendor's docs. Source.

What's in the skill

When it loads

Manages GenAI tuning jobs in Agent Platform. Use this to list, get, or cancel ongoing model tuning jobs. Don't use for fine-tuning models (use `agent-platform-tuning`), deploying models to endpoints (use `agent-platform-deploy`), or managing serving endpoints (use `agent-platform-endpoint-management`).

Sections in SKILL.md

Safety & Confirmation Tiers (CRITICAL)
Phase 0: Environment Setup
Workflow Decision Tree
Using the Python SDK
1. Listing Tuning Jobs (Tier R)
2. Getting Details for a Specific Job (Tier R)
3. Canceling a Job (Tier D)

Key facts

Author: Google, in Google's own google/skills repo (the marketplace owner is Google LLC)
License: Apache License 2.0 (repo LICENSE)
Version: 1.0.1 in the frontmatter, category AiAndMachineLearning
Installs on skills.sh: 8,917 as of 2026-11-02
GitHub stars: 21,764 for the whole google/skills repo, 2026-11-02
Last change to this skill: 2026-11-01
What is inside: 1 file: SKILL.md
Needs: Python 3, gcloud sign-in with Application Default Credentials, google-cloud-aiplatform, a Project ID and a region
Related skills: agent-platform-tuning, agent-platform-deploy, agent-platform-endpoint-management (named as the right skill for other jobs)
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass

What it includes

Tier R (list, get) runs freely. Tier D (cancel) needs your typed confirmation.
SKILL.md marks these rules as critical. Read-only calls need no confirmation. Cancelling needs a preview card and a typed "I confirm" or "Yes, cancel it" in a later turn.
Safety tiers
Lists gcloud sign-in steps and checks for the google-cloud-aiplatform package.
The steps are gcloud auth login, gcloud auth application-default login, and pip install google-cloud-aiplatform only when import vertexai fails. It says not to create a virtual environment.
Runs code · Phase 0: environment setup
Asks for the Project ID and region if missing, and never searches regions on its own.
Workflow decision tree
Prints the name, base model and state of every tuning job in a project and region.
Uses list_tuning_jobs on the GenAiTuningServiceClient in the Python SDK.
Try: “What tuning jobs do I have running in my project in us-central1?”
Runs code · List tuning jobs
Shows the name, base model, state and tuned model display name for one job ID.
Runs code · Get a tuning job
Shows a preview card, waits for your confirmation in a new turn, then cancels.
The card lists the full job resource name, the exact command and the effect: the job stops and in-progress training cannot be resumed.
Delete · Cancel a tuning job
On a 403, 404 or INVALID_ARGUMENT error it stops and asks for a valid Project or Job ID.
Error handling

From the official docs

Permissions

Medium risk

It can cancel a running Google Cloud tuning job, and that cannot be undone. Cancelling needs your typed confirmation, so check the job name on the preview card before you confirm.

What it can do

Read your data
Create and change things
Delete things
Read-only modeHow to turn it on

Watch out for

It can cancel a running tuning job
Cancelling halts in-progress training and it cannot be resumed. SKILL.md requires a preview card and your typed confirmation in a new turn. Check the full job name on the card. github.com
It runs under your Google Cloud login
The setup step lists gcloud auth login and gcloud auth application-default login. Every call runs with whatever those credentials can do. github.com
It may install a Python package
If import vertexai fails, SKILL.md runs pip install google-cloud-aiplatform, and tells the agent not to create a virtual environment. The package lands in your current Python environment. github.com

Technical details

Turning on read-only	Listing and getting jobs are Tier R (read-only) in SKILL.md. Cancel is a separate Tier D action that needs typed confirmation. Ask Claude to list or get only.
Talks to	{region}-aiplatform.googleapis.com
Scopes	Runs Python snippets that call the Agent Platform tuning API for your project and region Uses your gcloud Application Default Credentials

Questions

What does this skill do?

It makes Claude list, inspect and cancel GenAI tuning jobs in Agent Platform using the Python SDK. Claude asks for your Project ID and region, then runs short Python snippets.

Is it an official Google skill?

Yes. It is in Google's own google/skills repo, and the marketplace file there describes the repo as "A central repository of official skills and prescriptive plugins designed to provide a high-performance experience for building and deploying agentic applications across Google products and technologies."

How do I install it?

Use the install box above. The README command is npx skills add google/skills, which lets you pick skills, agents and scope. You also need Python and a Google Cloud sign-in.

Can Claude cancel a job without asking?

No. SKILL.md says cancellation needs explicit typed confirmation, and Claude must show a preview card with the job name, the exact command and the effect. It must wait for your confirmation in a new turn, even if you confirmed in advance.

Can it start or fine-tune a model?

No. The description says to use agent-platform-tuning for fine-tuning, agent-platform-deploy for deploying models to endpoints, and agent-platform-endpoint-management for serving endpoints.

What credentials does it use?

Your Google Cloud login. The setup step lists gcloud auth login and gcloud auth application-default login, and the Python snippets then call the regional aiplatform.googleapis.com endpoint with those credentials.

Is it safe?

Listing and getting jobs read your project, and cancelling stops a job for good: SKILL.md says in-progress training cannot be resumed. It may also install the google-cloud-aiplatform package into your Python environment. Our advice: use an account with only the permissions you need, and check the job ID on the preview card.

Information

updated Nov 2, 2026

Repository: google/skills
Folder: skills/cloud/agent-platform-tuning-management

google/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5

google/skills README · github.com
agent-platform-tuning-management SKILL.md · github.com
google-plugins marketplace.json · github.com
google/skills LICENSE · github.com
skills.sh listing · skills.sh