Gemini AI Studio to Agent Platform Migration
Guides Gemini API applications from Google AI Studio to Google Cloud Agent Platform.
The document describes IAM, budget alerts, and parts of the authentication flow, and places some setup actions with the user. However, it recommends the default service account, broad Cloud API access, creation and copying of a long-lived service-account key, and temporarily disabling an organization policy. Least privilege, sensitive-credential handling, data-flow disclosure, and rollback are materially incomplete, warranting a substantial deduction.
The main path includes gcloud commands, environment variables, SDK and ADK examples, OpenClaw configuration, and one stated OpenClaw version issue. However, numbering is inconsistent, the project-number retrieval command has semantic ambiguity, behavior depends on external product and version state, and abnormal-input and failure-feedback coverage is limited. Static review cannot exceed 10 points, so this receives a low-mid score.
Invocation scenarios, audience, and migration direction are fairly clear, covering local use, GCE, ADK, and OpenClaw. Boundaries for non-fit cases, regional availability, Chinese-language support, and mainland-China network reachability are not established, while the core flow depends on Google Cloud, OAuth, and several overseas services.
The document is organized into invocation conditions, comparison, migration steps, and additional resources. The repository README supplies Apache-2.0 licensing, installation, support, and active-development information. The skill itself lacks versioning, changelog, named maintenance responsibility, parameter contracts, and systematic troubleshooting, and some external-tool guidance may age quickly.
For a user who already has suitable Google Cloud access and clearly intends to migrate, the steps and configuration snippets may complete the core task. The workflow has substantial prerequisites involving payment, IAM, credentials, third-party CLI tooling, and OpenClaw versions, so environment-specific work remains significant. Static evidence provides no verified representative output, keeping the score at or below 7.
The skill includes several official-documentation links and auditable commands, but key claims about products, privacy, credits, models, and versions are delegated to external pages. There is no committed test suite, CI evidence, pinned dependency set, or independent reproduction record. Static review caps this at 5, and evidence coverage is thin.
- Do not follow the default path of creating or copying a long-lived service-account key; prefer short-lived credentials, workload identity, or a dedicated least-privilege service account.
- Do not disable organization policy or grant full cloud-platform access for troubleshooting without security approval, documented scope, and a tested recovery path.
- Review the source, version, permissions, and impact before running the remote curl script, installing Antigravity/OpenClaw, or restarting services.
- Revalidate billing, model names, privacy claims, regional endpoints, and OpenClaw compatibility before deployment; mainland-China reachability is not established by this skill.
What does this skill do, and when should you use it?
This skill guides the migration of Gemini API applications from Google AI Studio to Gemini Enterprise Agent Platform, formerly Vertex AI. It covers billing accounts, cloud credits, API enablement, authentication, and IAM. It also includes integration examples for the Google GenAI SDK, ADK, Antigravity CLI, and OpenClaw. It is intended for teams that want inference to share Google Cloud permissions, billing, and infrastructure.
It compares AI Studio and Agent Platform across endpoints, billing, privacy, security, throughput, and MLOps; provides gcloud commands for enabling the API, setting budget alerts, configuring ADC, and granting the aiplatform.user role; supplies environment variables, Python initialization code, and examples for ADK, Antigravity CLI, and OpenClaw; and describes how to restart and verify an OpenClaw connection.
- A developer is moving a Gemini API application from Google AI Studio to Google Cloud.
- A user wants Google Cloud Welcome Trial or other cloud credits to cover Gemini inference costs.
- A team needs Gemini inference to share IAM, billing, telemetry, Cloud SQL, or BigQuery infrastructure.
- A user runs ADK or OpenClaw on a Compute Engine VM and needs service-account access to Agent Platform.
What are this skill's strengths and limitations?
- Covers billing, credits, authentication, IAM, SDK configuration, and runtime verification in one migration guide.
- Includes copyable gcloud commands, environment variables, Python examples, and OpenClaw configuration.
- Addresses Google GenAI SDK, ADK, Antigravity CLI, and OpenClaw integrations.
- Requires a Google Cloud project, billing setup, and shell-based configuration.
- The OpenClaw section contains version and model restrictions that narrow its applicability.
- The source provides no test suite, automated migration script, or migration validation matrix.
How do you install this skill?
Run npx skills add google/skills, then select skills/cloud/agent-platform-migrate-from-ai-studio during the installation flow. The source does not document manually copying the folder into another client.
How do you use this skill?
In an Agent Skills-compatible client with this skill loaded, ask something such as “Help me migrate my Gemini API application from AI Studio to Agent Platform and configure IAM and billing.” Follow the guide to create billing and project resources, enable aiplatform.googleapis.com, configure ADC or a service account with roles/aiplatform.user, and set GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION, and GOOGLE_GENAI_USE_ENTERPRISE=TRUE. For OpenClaw, configure the google-vertex/ model prefix and restart the gateway.
How does this skill compare with similar options?
The skill explicitly compares Google AI Studio's Gemini Developer API with Agent Platform's Enterprise Gemini API across API endpoints, cloud-credit support, data privacy, IAM, security, SLAs, throughput, MLOps, and regional endpoints. Its practical distinction is that Agent Platform fits enterprise Google Cloud integration, while AI Studio remains the developer-oriented alternative.