Guide · Agent skill
How to use the Agent Platform Prompt Manager agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The agent-platform-prompt-management skill lets an AI agent work with prompt templates stored in Google's Agent Platform (formerly Vertex AI) for you: listing saved prompts, fetching one by ID or version, filling in its variables, creating new ones and deleting them. This guide is for developers and teams who already use Google Cloud and want to manage those prompts through a chat assistant. By the end you will know how to install the skill, what it needs, how to phrase requests, and how its confirmation steps protect your prompts.
Agent Platform Prompt Managergoogle/skills · Apache-2.0 Download the skill Quick start
- 1Download and unzip. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Authenticate with Google Cloud. Run gcloud auth login and gcloud auth application-default login with an account that can access your Agent Platform project.
- 3Check Python packages. Run the probe command and install google-cloud-aiplatform and google-genai only if the import fails.
- 4Ask for a list of prompts. Request a listing of managed prompts in your project and note the prompt IDs you need.
- 5Retrieve or fill a prompt. Ask for a prompt by ID or version, and optionally supply values for its variables.
- 6Approve or confirm changes. For a creation, reply Yes to the summary card. For a deletion, type your explicit confirmation in a new message.
What is an agent skill, and what does this one do?
An agent skill is a folder containing a SKILL.md instruction file and, in this case, supporting reference files. When the skill is installed, the agent reads those instructions and uses them when your request matches the skill's description. This skill covers managed prompts in Agent Platform and is meant for creating, listing, retrieving, versioning and deleting them.
The skill does this by running short Python snippets that use Google's SDK. It does not handle model training, model deployment to endpoints, or prompts stored in other tools.
- List prompts in a project and see their display names and IDs
- Retrieve a prompt by ID, or by a specific version
- Fill in a prompt's variables and print the assembled text
- Create a new prompt, after you approve a summary of it
- Delete a prompt, after you type an explicit confirmation
Install the skill from the ZIP
Download the ZIP for the skill from this page and unzip it. To make it available to Claude Code for all your projects, place the unzipped folder in ~/.claude/skills/. To make it available only inside one project, place it in that project's .claude/skills/ folder.
If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills
Set up Google Cloud authentication and Python
The skill needs Python 3, the Google Cloud CLI (gcloud), and access to an Agent Platform project with appropriate permissions. Authenticate with your Google Cloud account and set up Application Default Credentials, which the SDK uses to reach Agent Platform.
The skill's setup notes say to install the Python packages only if they are missing, and not to create a virtual environment. The probe command checks whether both packages import and installs them if not.
- Requires Python 3 and the Google Cloud CLI (gcloud)
- Requires an Agent Platform (Google Cloud) project with appropriate permissions
- Requires Python packages google-cloud-aiplatform and google-genai
gcloud auth login
gcloud auth application-default login
python3 -c "import vertexai, google.genai" \
|| pip install google-cloud-aiplatform google-genaiTrigger the skill with natural requests
You do not need special syntax. Describe what you want in plain language that mentions Agent Platform prompts. Include the details the agent will need: the project, the region, and the prompt ID where relevant. The skill's rules say the agent must ask you for any project ID, region, prompt ID or model it does not know, rather than guessing.
Read-only requests such as listing or fetching run without a confirmation step. Creating and deleting prompts pause for your approval first.
- "List the managed prompts in my Agent Platform project."
- "Get prompt 1234567890123456789 and show me its template text."
- "Get version 2 of that prompt and fill in user_name with Sam."
- "Create a prompt called Customer Support Greeting for a model I choose, with the template Hello {{user_name}}, how can I help."
- "Delete prompt 1234567890123456789."
What confirmations to expect
The skill sorts actions into tiers. Listing and getting prompts (read-only) run immediately. Creating a prompt is a mutating action: the agent shows you a card with the display name, target model and template text, and waits for a Yes or No before running anything. The target model must come from you; the skill does not supply a default.
Deleting a prompt is the strictest tier. The agent asks you to type an explicit confirmation, such as "I confirm", and waits for that in a new turn. Deletion is permanent. After a deletion, the agent lists prompts and checks that the ID is gone.
- Read-only (list, get): no confirmation
- Create: Yes/No approval of a summary card with parameters
- Delete: typed confirmation, in a separate turn, and irreversible
Safety notes: scripts and network access
This skill runs scripts and uses the network. Every operation calls Agent Platform with your authenticated credentials, so the account you log in with determines what the skill can read or change. Use a project and account where you are comfortable with those actions.
Deletions cannot be undone, so avoid running this skill unattended against production prompt assets. Review each confirmation card before approving it.
- Runs Python scripts and makes network calls to Agent Platform
- Uses your gcloud login and Application Default Credentials
- Check the project ID and region in every confirmation before approving
Tips for getting reliable results
Keep prompt IDs and version numbers handy, because the skill works with them directly. If you are unsure which prompt you mean, ask the agent to list prompts first and then pick one from the output.
When you create a prompt, state every parameter yourself, including the model. The skill is built so that parameters in the card trace back to what you said.
- List first, then get or delete, so you act on the ID you saw
- Name the model explicitly when creating a prompt
- Ask the agent to confirm a deleted prompt is gone after deletion
What people use it for
- Auditing which managed prompts exist in a project before a release
- Fetching a specific prompt version to compare its template text with another version
- Testing how a template renders by filling in its variables
- Creating a new prompt in a versioned prompt library after reviewing the parameters
- Removing an outdated test prompt after confirming its ID
Frequently asked questions
How do I install the agent-platform-prompt-management skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. In the Claude apps, upload it under Settings → Capabilities → Skills.
What are the requirements for this skill?
You need Python 3, the Google Cloud CLI (gcloud) with login and Application Default Credentials, the Python packages google-cloud-aiplatform and google-genai, and an Agent Platform (Google Cloud) project with appropriate permissions.
Is the Agent Platform Prompt Manager skill free?
The skill is published under the Apache-2.0 licence. The page does not state pricing for Google Cloud, which bills separately for Agent Platform usage, so check Google Cloud's own pricing for costs.
Can the skill delete prompts by accident?
Deletion requires you to type an explicit confirmation in a separate turn, and the agent is instructed not to run the deletion code before that reply. Deletions are permanent, so review the prompt ID before confirming.
Does this skill train or deploy models?
No. The skill explicitly excludes model training and deploying models to endpoints. It only handles managed prompts in Agent Platform.
Can it manage prompts stored in other tools?
No. It only works with Agent Platform managed prompts and is not meant for prompts kept in other tools.
What should I do if authentication fails?
Make sure you have run gcloud auth login and gcloud auth application-default login with an account that has access to your project. The skill needs both login and Application Default Credentials to reach Agent Platform.
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 developers already on Google Cloud Agent Platform who want to list, retrieve and manage versioned prompts through a chat assistant, with confirmation guardrails on creates and deletes. It is a poor fit if you lack a Google Cloud project with billing and permissions, need model training or deployment, or work with prompts stored elsewhere.
Download the skill