Guide · Agent skill
How to use the Agent Platform Endpoint Management agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
This guide explains what the Agent Platform Endpoint Management agent skill does, how to install it, and how to ask an agent to manage serving endpoints on Google Cloud's Agent Platform. It is written for readers who have not used agent skills before. By the end you will know the prerequisites, how to install the skill, how to phrase requests so the agent triggers it, and how its confirmation rules protect you from accidental changes.
Agent Platform Endpoint Managementgoogle/skills · Apache-2.0 Download the skill Quick start
- 1Check prerequisites. Confirm you have the gcloud CLI, a Google Cloud project and Agent Platform permissions such as aiplatform.admin or owner.
- 2Set up authentication. Run gcloud auth login and gcloud auth application-default login, then set your project with gcloud config set project.
- 3Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or your project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 4Ask a plain question. Ask about endpoints and name the region, for example listing endpoints in us-central1.
- 5Review any change before approving. For create or update, read the literal command the agent shows you and reply Yes only if it is what you want.
- 6Confirm deletes explicitly. For delete, type an explicit confirmation such as 'I confirm' in your reply. Make sure all models are undeployed first.
What does the skill make the agent do?
Endpoints on Agent Platform are logical serving hosts that provide a stable URL for online predictions. You must create an endpoint before you can deploy a model to it. Once this skill is installed, the agent knows how to list and describe existing endpoints and how to run gcloud commands that create, update or delete them.
- List endpoints in a region to find their IDs
- Describe a single endpoint to see its metadata and deployed models
- Create a new endpoint
- Update an endpoint's metadata, such as its display name or labels
- Permanently delete an endpoint
- Troubleshoot 403 permission, quota and resource busy errors
What do you need before you start?
The skill does not run anything on its own. The commands need a Google Cloud account with an active project, Agent Platform permissions (the skill refers to the aiplatform.admin or owner role), the Google Cloud SDK (gcloud CLI), and Application Default Credentials. Without these, the commands cannot run.
- Google Cloud SDK (gcloud CLI) installed
- Google Cloud account with an active project
- Agent Platform permissions, such as aiplatform.admin or owner
- Application Default Credentials configured
gcloud auth login
gcloud auth application-default login
gcloud config set project $PROJECT_IDHow do you install the skill?
Download the skill ZIP from this page and unzip it. Put the resulting skill folder in ~/.claude/skills/ to make it available to you across projects, or in a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills. The skill's files include SKILL.md, which contains the instructions the agent reads.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
How do you trigger the skill with a natural request?
You do not need a special command. Describe what you want in plain language that mentions endpoints. The skill is meant to activate when you need to create, list, describe, update or delete serving endpoints, or when you are troubleshooting endpoint permission, quota or resource busy errors.
Always name the region if you can. The skill tells the agent to ask for the region when you do not give one, and not to use 'global'.
- 'List the endpoints in us-central1.'
- 'Describe endpoint 1234567890 in us-central1.'
- 'Create an endpoint called fraud-scoring in us-east4.' (the agent will show you the exact command first)
- 'Rename endpoint 1234567890 to churn-v2 in us-central1.'
- 'I get a 403 when I list endpoints in europe-west4. What should I check?'
What happens when the agent makes a change?
The skill sorts actions into three tiers. Read-only actions (list, describe, get) run immediately with no confirmation. Create and update actions are reversible in the skill's classification, so the agent must show you the literal command, including all required flags, and wait for a 'Yes' in a later turn before running it. Delete is destructive and irreversible, so the agent asks for typed confirmation such as 'I confirm' or 'Yes, delete it' before doing anything else, and does not run pre-flight checks first.
The agent is not allowed to run a create, update or delete command in the same turn in which it asks for your confirmation. It has to stop and wait for your reply.
- Tier R (list, describe, get): runs immediately
- Tier M (create, update): shows the literal command, waits for Yes
- Tier D (delete): needs typed confirmation in a new turn
What do the core commands look like?
These are the commands the skill uses. Each one needs --region set to your location. The skill never uses 'global'. Pass your own values in place of the placeholders.
- Delete requires all models to be undeployed from the endpoint first. The skill's note says not to run describe until after you have typed confirmation.
- Create has no --asynchronous flag. The command waits and prints the new endpoint's resource name, whose last segment is the endpoint ID.
- Deploying or undeploying models is handled by the separate agent-platform-deploy skill.
gcloud ai endpoints list --region=$LOCATION_ID
gcloud ai endpoints describe $ENDPOINT_ID --region=$LOCATION_ID
gcloud ai endpoints create --region=$LOCATION_ID --display-name="my-endpoint"
gcloud ai endpoints update $ENDPOINT_ID --region=$LOCATION_ID --display-name="new-display-name"
gcloud ai endpoints delete $ENDPOINT_ID --region=$LOCATION_IDHow do you fix common errors?
The skill's troubleshooting notes cover three errors. A 403 Permission Denied usually means the account lacks the aiplatform.admin or owner role. A quota error means you should check the region's endpoint quota in the Cloud Console. A resource busy error during deletion often means models are still being undeployed from the endpoint.
- 403 Permission Denied: confirm the aiplatform.admin or owner role is assigned
- Quota Exceeded: check the region's endpoint quota in the Cloud Console
- Resource Busy on delete: check whether models are still being undeployed
What people use it for
- Listing endpoints in a region to find the IDs needed for other work
- Describing an endpoint to see which models are deployed on it
- Renaming an endpoint or changing its labels with a reviewed update command
- Creating a new endpoint with a display name before deploying a model through a separate skill
- Working through a 403, quota or resource busy error on an endpoint
Frequently asked questions
Is the Agent Platform Endpoint Management skill free to use?
The skill is published under the Apache-2.0 licence. The source does not cover Google Cloud pricing, so check Google Cloud's pricing for your project before running create or other commands that use cloud resources.
What do I need to install this agent skill?
You need the Google Cloud SDK (gcloud CLI), a Google Cloud account with an active project, Agent Platform permissions such as aiplatform.admin or owner, and Application Default Credentials set up with gcloud auth application-default login.
Where do I put the skill files?
Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. In the Claude apps, you can upload the ZIP under Settings → Capabilities → Skills.
Will the agent delete my endpoint without asking?
No. Delete requires explicit typed confirmation, such as 'I confirm' or 'Yes, delete it', and the agent must wait for you to reply in a new turn. Create and update also require a Yes after the literal command is shown.
Can this skill deploy or evaluate models?
No. The skill is for managing endpoints only. Deploying and undeploying models is covered by the separate agent-platform-deploy skill, and the skill explicitly excludes running model evaluations.
Why does my endpoint command fail with a 403 error?
A 403 Permission Denied error usually means your account lacks the required role. The skill's troubleshooting section says to make sure aiplatform.admin or owner is assigned.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits engineers who already run models on Google Cloud and want an agent to handle routine endpoint listing, describing and changes, with confirmation gates before any create, update or delete. It is not a good fit if you need to deploy or undeploy models, run evaluations, or do not yet have a Google Cloud project and gcloud access. Because deletes are irreversible, review every command the agent shows you before you approve it.
Download the skill