What does this skill do, and when should you use it?
This skill is for developers who manage prompts hosted in Agent Platform. It provides Python guidance for creating, listing, retrieving, versioning, assembling, and deleting managed prompts. Creation requires interactive confirmation, while deletion requires explicit typed confirmation in a later turn. Users must prepare Google Cloud authentication, a Python virtual environment, and the required SDK packages first.
It guides users through configuring a project and region with vertexai.init and using vertexai.preview.prompts to list prompts, retrieve prompts by ID, retrieve a specified version, and assemble template contents with variables. It requires confirmation of key parameters before creation and confirmation of the prompt ID before deletion. It also recommends double-curly-brace template variables, recording production version IDs, specifying a target model, and using the correct Dataset API metadata structure where applicable.
- A developer needs to inspect all managed prompts in a Google Cloud project.
- An application team needs to retrieve a prompt template and assemble it with runtime variables.
- A prompt engineer needs to create a managed prompt with a display name, template, and target model.
- A production maintainer needs to retrieve prompt versions and record their IDs.
- An administrator needs to remove an obsolete prompt with safeguards against accidental deletion.
How do you install this skill?
- Do not treat the documented confirmations as a substitute for least-privilege IAM or recovery controls; confirm resource scope and maintain an independent backup before deletion.
- The creation example has inconsistent placeholder syntax, and the SDK is preview with unpinned dependencies; verify the current API and versions before production use.
- Mainland-China reachability, Chinese interaction support, and cross-border data-processing risks are not documented.
- Shell / CLI
- Network access
- Local filesystem
PythonGoogle Cloud CLIgoogle-cloud-aiplatformgoogle-genai
Run npx skills add google/skills and select skills/cloud/agent-platform-prompt-management during installation. The repository is licensed under Apache-2.0 and bundles about 90 skills; this profile covers only the specified skill path.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/cloud/agent-platform-prompt-management" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
First run gcloud auth login and gcloud auth application-default login. Create and activate an environment with python3 -m venv ~/prompt_agent_venv and source ~/prompt_agent_venv/bin/activate, then install dependencies with pip install google-cloud-aiplatform google-genai. Activate the environment before every Python execution. Example trigger: List managed Agent Platform prompts for project PROJECT_ID in region LOCATION_ID. Creation and deletion require confirmation in a new turn; deletion code must not be provided before confirmation. Detailed creation snippets are referenced in references/create.md, but that file was not included in the source material.
What are this skill's strengths and limitations?
- Covers the main managed-prompt lifecycle: creation, listing, retrieval, version access, assembly, and deletion.
- Uses confirmation tiers to reduce accidental creation and irreversible deletion.
- Includes copyable Python SDK examples and environment setup commands.
- Clearly excludes model training, endpoint deployment, and non-Agent Platform prompt management.
- Requires Google Cloud authentication plus project and region configuration.
- The examples use `vertexai.preview.prompts`, but the source does not document its stability or compatibility range.
- Detailed creation code is not included in the provided material.
- The source provides no test suite, cost guidance, or comprehensive failure-handling documentation.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Agent Platform Prompt Manager this page ✓ Google · Official | 47 · Use with care | ★ 21k | 13d ago | Apache-2.0 |
| Agent Platform Tuning Manager ✓ Google · Official | 47 · Use with care | ★ 21k | 13d ago | Apache-2.0 |
| Gemini API Enterprise Development Guide ✓ Google · Official | 49 · Use with care | ★ 21k | 13d ago | Apache-2.0 |
| Gemini Enterprise Skill Registry ✓ Google · Official | 43 · Not recommended | ★ 21k | 13d ago | Apache-2.0 |
| Microsoft Foundry Python Projects Skill ✓ Microsoft · Official | 47 · Use with care | ★ 2.7k | 3mo ago | MIT |
How did FollowSkills review this skill?
The evidence shows interactive confirmation and same-turn restrictions for creation and deletion, plus ADC setup guidance, providing basic change control. Points are deducted because least-privilege IAM, data flows, sensitive-content handling, dependency integrity, and post-deletion recovery are not specified; the authentication commands may also alter the user's environment.
The evidence covers list, get, create, delete, versioning, placeholders, and some idempotency guidance. Points are deducted for the absence of tests, abnormal-input handling, or diagnosable failure feedback; placeholder syntax differs between the main guide and creation example, while preview APIs and unpinned dependencies add uncertainty. Static review caps this at 10.
The metadata clearly defines prompt-management scenarios and excludes training, endpoint deployment, and non-Agent Platform prompts. Points are deducted because input/output contracts, precise trigger boundaries, atypical failure cases, Chinese-language support, and mainland-China network reachability are not documented.
The evidence shows layered documentation with a main guide and creation reference, including installation, examples, safety requirements, licensing, and a repository support path. Points are deducted for lacking a skill-specific versioning policy, changelog, named maintenance responsibility, explicit update path, FAQ, systematic troubleshooting, and pinned dependency versions.
The evidence provides Python snippets for listing, retrieving, creating, and deleting managed prompts, so the core task is broadly covered. Points are deducted because the code was not executed or independently verified, creation guidance is brief, and complete update/version-management flows, error handling, and directly runnable parameter examples are missing. Static review caps this at 7.
The evidence includes source snippets, operational constraints, README context, and a license, but no committed tests, CI coverage, or third-party execution evidence. Points are deducted because SDK behavior, version compatibility, and example correctness cannot be independently reproduced from the supplied files; static review caps this at 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →