google-cloud-recipe-foundation-builder

SKILL.md

Google Cloud Recipe: Foundation Builder

[!WARNING] This skill is currently in a preview state. It will deploy a secure
foundation, but does not have all advanced features. Users who want more
options should visit
Google Cloud Setup.

This skill guides the setup of a secure, enterprise-grade Google Cloud landing
zone foundation. It establishes baseline security controls, organizes the
initial resource hierarchy, and configures centralized audit logging and
cross-environment monitoring.

Overview

The recipe provisions the following core components at the organization root:

Security Guardrails: Enforces 17 baseline Google Cloud Organization
Policies to secure the environment (13 Boolean, 4 List constraints).

Resource Hierarchy: Establishes 4 folders (Common, Production,
Non-Production, Development) and provisions corresponding projects
sequentially with globally unique ID prefixes (logging-, prod-,
non-prod-, dev- followed by a shared suffix).

Billing & API Enablement: Links all projects to your billing account and
activates critical logging/monitoring services.

Centralized Logging & Monitoring: Deploys a global, centralized log
bucket with 30-day retention, configures an organization-wide audit log
sink, and sets up a cross-environment metrics scope.

Clarifying Questions

Before executing this recipe, the agent must gather the following details:

Organization ID: Run gcloud organizations list to retrieve the
available organizations, present them to the user, and ask them to select
the target Organization ID.

Billing Account ID: Run gcloud billing accounts list --filter=open=true to retrieve only the active (open) billing accounts,
present them to the user, and ask them to select the active Billing
Account ID.

Project ID Suffix: Ask if the user has a preferred prefix or target
suffix for Project IDs (default uses prefix + a shared random 8-character
string, e.g., prod-ab12cd34).

Log Bucket Region: Ask for the target region for resources if they want
to override the default global log bucket location.

Prerequisites

Ensure the following prerequisites are met before beginning the deployment:

GCP Identity: You must have a Google Cloud Organization resource set up.

Administrative IAM Roles: The identity executing these commands must
hold the required administrative permissions. If any step fails with a
Permission Denied error, the agent will attempt to self-remediate by
granting the corresponding recommended role as detailed in Phase 2: Error
Recovery & Lazy Role Remediation Strategy.

Tools: The gcloud CLI must be installed, authorized with the above
identity, and configured for use.

Steps to Complete the Recipe

Phase 1: Pre-flight Confirmation

Identify the target organization and obtain explicit user approval before making
changes.

Identify and Discover Organization: Verify the target organization. If
only the display name is known, list organizations to find the ID. Then,
retrieve the organization metadata to dynamically calculate the Directory
Customer ID and Domain Name:

# List to find ID if needed
gcloud organizations list

# Describe the organization to retrieve metadata
gcloud organizations describe [ORGANIZATION_ID]

Calculate values:

Domain Name ([ORG_NAME] / [YOUR_DOMAIN]): Use the displayName
value from the output (e.g., my-business.com).

Customer ID ([DIRECTORY_CUSTOMER_ID]): Use the
owner.directoryCustomerId value from the output (e.g., C01234567).

Present Blueprint Summary: Present the exact details of the blueprint to
the user and request confirmation to proceed:

Proposed Foundation Deployment Summary for Organization: [ORG_NAME]
([ORGANIZATION_ID])

Security: Enforce 17 baseline Organization Policies (13 Boolean, 4
List).

Folders: Create 4 folders sequentially (Common, Production,
Non-Production, Development).

Projects: Create 4 projects sequentially with unique IDs
(logging-[SUFFIX], prod-[SUFFIX], non-prod-[SUFFIX],
dev-[SUFFIX]).

Billing: Link all projects to Billing Account
[BILLING_ACCOUNT_ID].

APIs: Enable Logging and Monitoring APIs on the central project.

Centralized Logging: Deploy a global log bucket
[ORG_NAME]-logging (30-day retention), configure an
organization-level sink
[ORGANIZATION_ID]-logbucketsink-[RANDOM_HEX], and establish
cross-project metrics scopes.

Do you wish to proceed with this deployment? (Yes/No)

[!IMPORTANT] Pause execution and wait for explicit user approval before
moving to Phase 2. If the user declines, abort the operation.

Phase 2: Error Recovery & Lazy Role Remediation Strategy

To ensure deployment continues smoothly on clean organizations without requiring
complex upfront permission checks (which require a quota project), the agent
must adopt a "lazy recovery" approach.

Instead of testing permissions in advance, the agent will attempt to execute
each step in the recipe. If a step fails with a Permission Denied error, the
agent will attempt to self-remediate by granting the corresponding recommended
administrative group of roles to the deployment identity and retrying the
operation.

[!IMPORTANT] When asked about pre-deployment readiness, prerequisites, or
checks to run, the agent must explicitly explain that the landing zone
deployment adopts a lazy role remediation strategy rather than upfront
testing, detailing all of the following in its response: 1. Confirm it will
execute deployment commands directly, catching any Permission Denied
errors. 2. Confirm it will attempt to self-remediate on failure by running the
exact commands gcloud organizations add-iam-policy-binding or gcloud billing accounts add-iam-policy-binding to grant the entire administrative
group of roles to the active identity, and then retry the failed deployment
command. 3. List the core administrative groups that it will attempt to grant
(Organization Admin Group, Billing Admin Group, and Security Admin Group)
mapped to their key roles. 4. Confirm it will halt execution and request
manual administrator intervention if the self-remediation grant command fails.

Remediation Protocol

For any command that fails due to missing permissions:

Identify Required Admin Group: Determine which administrative group is
responsible for the failed action. Refer to the
Administrative IAM Reference
for details.

Attempt Self-Remediation: Grant all roles belonging to that
administrative group to the active authenticated account sequentially. Refer
to the
Administrative IAM Reference Remediation Guide
for the copy-pasteable script commands:

For Organization/Folder level failures (Org Admin Group or Security
Admin Group): Run gcloud organizations add-iam-policy-binding
sequentially for each role in the group.

For Billing level failures (Billing Admin Group): Run gcloud billing accounts add-iam-policy-binding sequentially for each role in
the group.

Halt on Remediation Failure:

If the grant commands succeed, immediately retry the failed deployment
command.

If any of the grant commands fail (e.g., due to lack of setIamPolicy
admin rights), halt execution and instruct the user to ask their
Organization/Billing Administrator to manually grant the entire
administrative group of roles.

Phase-Specific Remediation Mapping

Phase 3: Security Guardrails (Org Policies):

If gcloud org-policies set-policy fails: Attempt to grant the entire
Organization Admin Group (9 roles) at the organization level.

Phase 4: Resource Hierarchy (Folders & Projects):

If gcloud resource-manager folders create or gcloud projects create
fails: Attempt to grant the entire Organization Admin Group (9
roles) at the organization level.

Phase 4: Billing Link:

If gcloud billing projects link fails: Attempt to grant the entire
Billing Admin Group (3 roles) at the billing account level, and
ensure the active identity is granted the Organization Admin Group
(which contains roles/billing.user) at the organization level.

Phase 5: Centralized Logging & Monitoring:

If gcloud logging sinks create fails at org level: Attempt to grant
the entire Logging/Monitoring Admin Group (2 roles:
roles/logging.admin, roles/monitoring.admin) and the Security
Admin Group (9 roles) at the organization level.

Phase 3: Security Guardrails (Org Policies)

Apply 17 baseline security controls at the organization root.

[!CAUTION] Applying iam.allowedPolicyMemberDomains first can lock out the
deployment identity if it resides in an unallowed domain. Ensure the
deployment identity is safe before enforcing this policy.

Generate the YAML configuration files for the 17 policies. Refer to the
Organization Policies Reference for the exact
YAML templates for both Boolean and List constraints.

Apply each organization policy sequentially using the gcloud org-policies
tool:

gcloud org-policies set-policy [POLICY_FILE_NAME].yaml

Phase 4: Resource Hierarchy

1. Folder Creation

Check if target folders exist to avoid duplication. The agent must check for all
4 folders: for any folder that already exists (e.g., if Common or Production
are already present), the agent must locate and reuse them; for any folder that
is missing (e.g., if Non-Production or Development are not present), the
agent must proceed to sequentially create them:

[!IMPORTANT] When explaining how existing resources (folders and projects) are
handled to prevent duplication, the agent must explicitly name the
remaining missing folders (Non-Production and Development) and confirm
that it will proceed to sequentially create only these missing folders and
projects.

# Check and Create "Common" Folder
gcloud resource-manager folders list --organization=[ORGANIZATION_ID] --filter="display_name=Common"
# If not present:
gcloud resource-manager folders create --display-name="Common" --organization=[ORGANIZATION_ID]

# Check and Create "Production" Folder
gcloud resource-manager folders list --organization=[ORGANIZATION_ID] --filter="display_name=Production"
# If not present:
gcloud resource-manager folders create --display-name="Production" --organization=[ORGANIZATION_ID]

# Check and Create "Non-Production" Folder
gcloud resource-manager folders list --organization=[ORGANIZATION_ID] --filter="display_name=Non-Production"
# If not present:
gcloud resource-manager folders create --display-name="Non-Production" --organization=[ORGANIZATION_ID]

# Check and Create "Development" Folder
gcloud resource-manager folders list --organization=[ORGANIZATION_ID] --filter="display_name=Development"
# If not present:
gcloud resource-manager folders create --display-name="Development" --organization=[ORGANIZATION_ID]

2. Project Creation and Billing Link

Check if target projects already exist in the folders by matching their display
names. If not present, generate a shared 8-character random suffix (e.g.,
ab12cd34) and create the projects sequentially, linking billing and enabling
APIs immediately:

# Check if "central-logging-monitoring" project exists in Common folder
gcloud projects list --filter="parent.id=[COMMON_FOLDER_ID] AND parent.type=folder AND name=central-logging-monitoring"

# If not present: Create, link billing, and enable APIs
gcloud projects create logging-[SUFFIX] --name="central-logging-monitoring" --folder=[COMMON_FOLDER_ID]
gcloud billing projects link logging-[SUFFIX] --billing-account=[BILLING_ACCOUNT_ID]
gcloud services enable compute.googleapis.com logging.googleapis.com monitoring.googleapis.com --project=logging-[SUFFIX]

# Check if "production" project exists in Production folder
gcloud projects list --filter="parent.id=[PRODUCTION_FOLDER_ID] AND parent.type=folder AND name=production"

# If not present: Create, link billing, and enable APIs
gcloud projects create prod-[SUFFIX] --name="production" --folder=[PRODUCTION_FOLDER_ID]
gcloud billing projects link prod-[SUFFIX] --billing-account=[BILLING_ACCOUNT_ID]
gcloud services enable compute.googleapis.com run.googleapis.com container.googleapis.com artifactregistry.googleapis.com firestore.googleapis.com pubsub.googleapis.com aiplatform.googleapis.com cloudaicompanion.googleapis.com apphub.googleapis.com designcenter.googleapis.com discoveryengine.googleapis.com iam.googleapis.com config.googleapis.com cloudbuild.googleapis.com cloudasset.googleapis.com cloudkms.googleapis.com cloudresourcemanager.googleapis.com --project=prod-[SUFFIX]

# Check if "non-production" project exists in Non-Production folder
gcloud projects list --filter="parent.id=[NON_PRODUCTION_FOLDER_ID] AND parent.type=folder AND name=non-production"

# If not present: Create, link billing, and enable APIs
gcloud projects create non-prod-[SUFFIX] --name="non-production" --folder=[NON_PRODUCTION_FOLDER_ID]
gcloud billing projects link non-prod-[SUFFIX] --billing-account=[BILLING_ACCOUNT_ID]
gcloud services enable compute.googleapis.com run.googleapis.com container.googleapis.com artifactregistry.googleapis.com firestore.googleapis.com pubsub.googleapis.com aiplatform.googleapis.com cloudaicompanion.googleapis.com apphub.googleapis.com designcenter.googleapis.com discoveryengine.googleapis.com iam.googleapis.com config.googleapis.com cloudbuild.googleapis.com cloudasset.googleapis.com cloudkms.googleapis.com cloudresourcemanager.googleapis.com --project=non-prod-[SUFFIX]

# Check if "development" project exists in Development folder
gcloud projects list --filter="parent.id=[DEVELOPMENT_FOLDER_ID] AND parent.type=folder AND name=development"

# If not present: Create, link billing, and enable APIs
gcloud projects create dev-[SUFFIX] --name="development" --folder=[DEVELOPMENT_FOLDER_ID]
gcloud billing projects link dev-[SUFFIX] --billing-account=[BILLING_ACCOUNT_ID]
gcloud services enable compute.googleapis.com run.googleapis.com container.googleapis.com artifactregistry.googleapis.com firestore.googleapis.com pubsub.googleapis.com aiplatform.googleapis.com cloudaicompanion.googleapis.com apphub.googleapis.com designcenter.googleapis.com discoveryengine.googleapis.com iam.googleapis.com config.googleapis.com cloudbuild.googleapis.com cloudasset.googleapis.com cloudkms.googleapis.com cloudresourcemanager.googleapis.com --project=dev-[SUFFIX]

[!NOTE] Agentic Parallelism Option: While the manual runbook enforces
sequential project execution to avoid terminal race conditions, an AI agent
with multi-agent orchestration capability may optionally spawn subagents to
provision the 4 projects in parallel once folder IDs are resolved.

Phase 5: Centralized Logging and Monitoring

Configure centralized audit logging and cross-project monitoring scope in the
logging-[SUFFIX] project.

Refer to the
Centralized Logging and Monitoring Reference
for the detailed step-by-step commands to:

Create the central log bucket.

Create the organization-wide log sink.

Grant required IAM permissions to the log sink.

Configure the cross-project monitoring metrics scope.

Validation Logic & Checklist

Evaluate the deployment against the following verification checks:

 Security Policies: Run gcloud org-policies list --organization=[ORGANIZATION_ID] and verify all 17 target policies are
enforced or correctly configured.

 Resource Folders: Verify folders Common, Production,
Non-Production, and Development exist under the organization root.

 Billing Linkage: Run gcloud billing projects list and assert that
all 4 newly created projects are linked to your billing account.

 Log Bucket & Retention: Verify the log bucket [ORG_NAME]-logging
exists in project logging-[SUFFIX], is located in global, and has a
retention period of exactly 30 days.

 Log Sink Routing: Run gcloud logging sinks describe at the
organization level and confirm the sink routes cloud audit logs to the
global bucket and holds standard writerIdentity credentials.

 Metrics Scope Linkage: Run gcloud beta monitoring metrics-scopes describe and assert that the dev, non-prod, and prod projects appear
in the monitored list of the central logging project.

Links

Google Cloud Resource Hierarchy Documentation

Google Cloud Organization Policies Overview

Centralized Audit Logging Best Practices

Cloud Monitoring Metrics Scopes Configuration

Gcloud Logging Sinks CLI Reference

Google Cloud Landing Zones Guide

Google Cloud Security Foundations Blueprint