Deploy open models to Agent Platform: cost-checked, confirmation-gated
Deploy Model Garden open models (or 1P tuned models) to Agent Platform endpoints with tiered confirmations — read-only is free to run, deploys need a cost-checked dry-run card, deletes need typed confirmation
- What
- Deploy Model Garden open models (or 1P tuned models) to Agent Platform endpoints with tiered confirmations — read-only is free to run, deploys need a cost-checked dry-run card, deletes need typed confirmation
- Cost
- Free
- Needs
- a Google Cloud account with billing enabled (deploying provisions REAL cloud compute with REAL hourly costs — nothing here is free); gcloud CLI authenticated to the project; the skill is guidance the agent follows, not software to install
- Install
- Copy the installer prompt below into your Muse — your agent does the rest.
Curated by Skill Harbor — Google's official skill for deploying open models and custom weights from Model Garden to Agent Platform endpoints: discover deployable models via `gcloud ai model-garden models list`, check supported machine types and accelerators per model, then deploy with the `:deploy` API. Its defining feature is a strict safety-and-confirmation tier system: read-only actions (list, describe) run freely; mutating actions (deploy, undeploy) require an explicit dry-run confirmation card showing the model ID, project, region, machine type, endpoint display name and — critically — an estimated hourly cost that must be computed (via the bundled calculate_cost.py script or the published pricing page, never invented); destructive actions (delete) require typed confirmation. Includes status checks (single describe per turn — no sleep loops), test-prediction verification via the endpoint's dedicated DNS, undeploy/cleanup guidance, a 1P tuned-model cross-region copy guide, quota-troubleshooting playbooks, and a cost-pushback/region-failover renegotiation flow. Honest caveats: deploying provisions REAL Google Cloud compute — a Google Cloud account with billing enabled is required and every deploy incurs real hourly costs (undeploy when done or charges continue); the skill never names model versions from memory — it always re-queries the live catalog. Apache-2.0 licensed. Skill Harbor never reviews the code, review it yourself before use. Discovered via skills.sh.
Version:
Install
Prerequisites: a Google Cloud account with billing enabled (deploying provisions REAL cloud compute with REAL hourly costs — nothing here is free); gcloud CLI authenticated to the project; the skill is guidance the agent follows, not software to install Install "Deploy open models to Agent Platform: cost-checked, confirmation-gated" for me. It teaches my agent Google's deployment workflow: discover deployable Model Garden models with the live catalog (never name model versions from memory), check machine-type/accelerator support per model, compute the hourly cost with the bundled script (never invent a number), present a cost-checked dry-run confirmation card before any deploy, run single status checks (no polling loops), verify with a test prediction, and undeploy/clean up afterwards — including the 1P tuned-model cross-region copy guide. Apache-2.0 licensed. Repository: https://github.com/google/skills/blob/main/skills/cloud/agent-platform-deploy/SKILL.md 1. Fetch the SKILL.md file (and any helper files) from the repository path into a temporary folder and summarize what it does in one or two sentences. 2. Safety check: review the SKILL.md and scripts for anything suspicious (unexpected network calls, shell commands, credential harvesting). This repo should contain zero secrets in code, credentials only via the secure vault, allowed hosts declared in the SKILL.md. Verify that holds here; STOP on any red flag and tell me. 3. Install it as a skill: copy SKILL.md and its helper files into the agent's skills directory, in a folder named "agent-platform-deploy". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (e.g. authenticate gcloud (gcloud auth login) and set the target project/region; remember every deploy incurs real hourly cloud costs and must be undeployed when done). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.
Questions
How do I install a build?
Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the build up for you — no manual configuration.
Where does my money go?
Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.
What does the ✓ next to a creator’s name mean?
It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a build before installing it.