Google ADK Agent Deployment Pilot

Produces a step-by-step deployment guide for moving Python agents written with the Google Agent Development Kit (ADK) into production. It builds on the agents-cli tool, wrapping Terraform, Docker, and the deployment pipeline into a single tested pipeline.

What it does: Helps you pick the right deployment target (Agent Runtime / Cloud Run / GKE) for your needs; provides concrete commands to enable the required GCP APIs, manage API keys with Secret Manager, set up service account (app_sa / cicd_runner_sa) IAM bindings, run a HITL-approved dev deployment, set up a CI/CD pipeline, and test the deployed agent. It resolves common 403/503 errors and rollback paths (git-based, Cloud Run traffic shifting, kubectl rollout undo).

When to use it: in cases like "deploy my agent," "deploy the ADK agent," "set up CI/CD," "configure secrets," "fix a deployment issue." All three deployment targets are valid production options; they have different trade-offs.

Output: a Markdown guide with a decision matrix, runnable gcloud / agents-cli / kubectl command blocks, an IAM error-resolution table, and rollback instructions. Critical rule: agents-cli deploy is never run without explicit human approval.