Google Agent Platform Endpoint Management Skill for Claude Code: What It Does and How to Install It
Makes Claude manage Google Cloud Agent Platform serving endpoints with gcloud: list, describe, create, update and delete, with confirmation before changes.
- Installs
- 8.5K
- on skills.sh
- Stars
- 21.8K
- on GitHub
- Updated
- Nov 2026
- last repo commit
The skill sorts every action into three tiers. Reading (list, describe) runs right away. Creating or updating needs a Yes/No card that shows the exact gcloud command, and Claude must wait for your reply. Deleting needs you to type a confirmation such as "I confirm", and Claude must ask before it does any checks.
It needs the gcloud CLI, a Google Cloud project and a region (it must not use global). It manages endpoints only. Deploying models to an endpoint belongs to the agent-platform-deploy skill, and the description says not to use this one for deployment or evaluations.
Good for
- Developers who host models on Google Cloud Agent Platform endpoints
- Platform engineers who create or clean up serving endpoints
- Anyone debugging endpoint permission, quota or resource busy errors
Install agent-platform-endpoint-management
Claude Code
- 1
Copy this prompt.
Install the agent-platform-endpoint-management skill in Claude Code. Run this command: npx skills add https://github.com/google/skills --skill agent-platform-endpoint-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, gcloud CLI and a Google Cloud project.
Pick one route. The README command npx skills add google/skills lets you pick skills when prompted; the command here installs only agent-platform-endpoint-management. No plugin in the repo marketplace holds this skill. The skill runs gcloud, so there is no Claude app upload route.
Built from the package data we hold. Not yet tested by us.
What's in the skill
When it loads
Manages Agent Platform serving endpoints. Use when you need to create, list, describe, update, or delete serving endpoints for model deployment on Agent Platform. Also use when troubleshooting endpoint permission, quota, or resource busy errors. Don't use for deploying models to endpoints or for running model evaluations.
Sections in SKILL.md
- Overview
- Safety & Confirmation Tiers (CRITICAL)
- Phase 0: Environment Setup
- 1. Listing Endpoints (Tier R)
- 2. Describing an Endpoint (Tier R)
- 3. Creating an Endpoint (Tier M)
- 4. Updating an Endpoint (Tier M)
- 5. Deleting an Endpoint (Tier D)
- 6. Traffic Splitting (Tier M)
- Troubleshooting
Key facts
- Repo
- google/skills
- Installs on skills.sh
- 8,499 (Nov 2, 2026)
- GitHub stars
- 21,790 for the whole repo (Nov 2, 2026)
- Last repo commit
- Nov 1, 2026
- Audits on skills.sh
- Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass
- What it includes
- 1 file: SKILL.md (no scripts)
What it includes
Runs gcloud auth login, application-default login and sets the active project.
Runs code ·
Environment setupgcloud ai endpoints list --region=..., runs without confirmation.
Optional --limit, --page-size and --page-token.
Try: “List my endpoints in us-central1”
Runs code ·
List endpoints (Tier R)gcloud ai endpoints describe shows metadata, including deployedModels.
Runs code ·
Describe an endpoint (Tier R)gcloud ai endpoints create with a display name, after a Yes/No card.
The card must show the literal command. Claude may not run it in the same turn.
Runs code ·
Create an endpoint (Tier M)gcloud ai endpoints update changes the display name or labels, after a Yes/No card.
SKILL.md says to check the endpoint exists first.
Runs code ·
Update an endpoint (Tier M)gcloud ai endpoints delete removes an endpoint for good, after typed confirmation.
All models must be undeployed first.
Delete ·
Delete an endpoint (Tier D)Points to the agent-platform-deploy skill for deploying and undeploying models.
The section holds only a comment about gcloud ai endpoints deploy-model.
Traffic splitting (Tier M)Covers 403 permission errors, quota exceeded and resource busy on delete.
SKILL.md: for resource busy, check whether models are still being undeployed.
Troubleshooting
From the official docs
Permissions
High risk
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 delete endpoints permanently
gcloud ai endpoints delete is in the skill. SKILL.md requires typed confirmation, asked immediately and before any pre-flight checks, with Claude waiting for your reply in a new turn. Read the project and endpoint ID in the request before you confirm. github.com
It runs gcloud as you
Commands run with your signed-in Google Cloud credentials and active project. Check which project gcloud points at before you start. github.com
Create and update change resources in your project
Create adds a serving endpoint and update changes its display name or labels. SKILL.md requires a Yes/No card with the exact command for both, so review that command each time. github.com
Troubleshooting suggests broad roles
For a 403 the skill says to ensure the aiplatform.admin or owner role is assigned. Prefer the narrowest role that works. github.com
Technical details
| Turning on read-only | List and describe are the read-only tier and run without confirmation. Ask only for those, or use a read-only IAM role. |
|---|---|
| Sign-in and keys | Uses your gcloud sign-in and Application Default Credentials. SKILL.md names no API key. |
| Scopes |
|
Questions
What does the agent-platform-endpoint-management skill do?
It makes Claude list, describe, create, update and delete Google Cloud Agent Platform serving endpoints with gcloud ai endpoints commands. It also covers permission, quota and resource busy errors.
How do I install it?
Use the install box at the top of this page. The README documents the skills CLI for the google/skills repo and says you can pick skills when prompted. No plugin in the repo marketplace holds this skill.
Will it delete an endpoint without asking?
SKILL.md says deletion needs explicit typed confirmation, asked immediately and before any checks, and Claude must wait for your reply in a new turn before running it. Create and update need a Yes/No card showing the exact command, also with no execution in the same turn.
Can it deploy a model?
No. The description says not to use it for deploying models to endpoints or running evaluations. SKILL.md points to the agent-platform-deploy skill for deploying and undeploying.
Does it need a Google Cloud account?
Yes. Setup runs gcloud auth login and gcloud auth application-default login, then sets the project. Every command also needs a region, and SKILL.md says not to use global.
Does it work in the Claude app?
The install box has no Claude app upload route. The skill runs gcloud on your machine.
Is it safe?
It can create, change and permanently delete endpoints in your project. Use a test project first and read each command in the confirmation card.
Is it an official Google skill?
It lives in Google's own google/skills repo, whose README says it holds Agent Skills for Google products.
Information
updated Nov 2, 2026
- Repository
- google/skills
- Folder
skills/cloud/agent-platform-endpoint-management
Sources
Docs, issue trackers and discussions this page is built from.
Show all 3Hide
- google/skills README · github.com
- agent-platform-endpoint-management SKILL.md · github.com
- google/skills marketplace.json · github.com