deploy-blueprint

name	deploy-blueprint
description	Deploy an Aviatrix blueprint to your environment with guided, multi-step orchestration. Handles both simple single-layer and complex multi-layer blueprints.
argument-hint	[blueprint-path] [--plan-only] [--layer <layer-name>]
disable-model-invocation	true
context	fork
agent	general-purpose
allowed-tools	["Bash(terraform *)","Bash(aws *)","Bash(kubectl *)","Bash(helm *)","Bash(gh *)","Bash(find *)","Bash(source *)","Task","AskUserQuestion","Read","Grep","Glob","Write"]

Deploy Blueprint

Deploy an Aviatrix blueprint to your environment with guided, multi-step orchestration.

Usage

/deploy-blueprint [blueprint-path] [--plan-only] [--layer <layer-name>]

Options:

• --plan-only: Review deployment plan without applying changes
• --layer <layer>: Deploy specific layer only (e.g., network, clusters, nodes)

If no path is provided, prompts for blueprint selection.

Prerequisites

This skill requires:

• Aviatrix Control Plane accessible (Controller and optionally CoPilot)
• Cloud provider credentials configured (AWS CLI, Azure CLI, or gcloud)
• Terraform v1.5+ installed
• Valid terraform.tfvars with appropriate values

What This Skill Does

1. Blueprint Selection - Choose which blueprint to deploy
2. README Review - Display key information from documentation
3. Prerequisites Check - Verify tools, access, and configuration
4. Environment File Setup - Create/verify credentials file
5. Deployment Planning - Analyze architecture and create execution plan
6. Orchestrated Deployment - Execute deployment with subagents for efficiency
7. Validation - Verify deployment success
8. Post-Deployment - Provide next steps and access information

Deployment Workflow

Phase 1: Blueprint Selection

If no blueprint path provided, display available blueprints and prompt for selection.

Phase 2: README Review

Display key sections from the blueprint README:

• Description and architecture summary
• Estimated cost and deployment time
• Link to full README

Ask for confirmation to continue.

Phase 3: Prerequisites Check

Verify all required tools and access:

• Terraform version
• Cloud CLI tools
• Configuration files
• Controller connectivity

Interactively prompt for:

• Controller URL and credentials
• Cloud account information
• Blueprint-specific variables

Phase 4: Environment File Setup

Create or verify environment file for credentials:

# Create .env.blueprint file
cat > .env.blueprint <<EOF
export AVIATRIX_CONTROLLER_IP="${controller_ip}"
export AVIATRIX_USERNAME="${username}"
export AVIATRIX_PASSWORD="${password}"
export AWS_REGION="${aws_region}"
EOF

# Add to .gitignore
echo ".env.blueprint" >> .gitignore

# Source the file
source .env.blueprint

Phase 5: Deployment Planning

Analyze blueprint architecture:

• Detect if single-layer or multi-layer
• For multi-layer: identify dependencies and deployment order
• Run terraform plan for each layer to count resources
• Display deployment summary with time estimates

Ask for confirmation to proceed.

Phase 6: Orchestrated Deployment

For multi-layer deployments:

Layer 1: Sequential deployment (foundation)
  └─ Deploy network layer

Layer 2: Parallel deployment (if multiple subdirectories)
  ├─ Spawn subagent for clusters/frontend
  └─ Spawn subagent for clusters/backend

Layer 3: Parallel deployment (if multiple subdirectories)
  ├─ Spawn subagent for nodes/frontend
  └─ Spawn subagent for nodes/backend

Use the Task tool with subagent_type=Bash to spawn deployment agents for parallel execution.

For single-layer deployments:

Use a single agent for straightforward deployment.

Phase 7: Validation

Verify deployment success:

• Check gateway status in Controller
• Verify EKS clusters (if applicable)
• Run connectivity tests
• Validate against expected outputs

Phase 8: Post-Deployment

Provide access information:

• Kubernetes cluster access commands
• Controller/CoPilot URLs and navigation
• Sample application deployment instructions
• Next steps from README
• Destroy instructions (in reverse order)

Instructions for Claude

When this skill is invoked:

1. Blueprint Selection

# If no path provided, list available blueprints
find blueprints/ -maxdepth 1 -type d -not -name ".*" -not -name "blueprints"

# Read README.md from selected blueprint
# Extract: title, description, estimated cost, deployment time, architecture type

2. Prerequisites Check

# Check required tools
terraform --version
aws --version
kubectl version --client

# Check for configuration files
test -f terraform.tfvars || test -f terraform.tfvars.example

Use AskUserQuestion to collect:

• Aviatrix Controller URL
• Aviatrix credentials
• Cloud provider information
• Blueprint-specific variables

3. Environment File Creation

# Check if .env.blueprint exists
if [ ! -f .env.blueprint ]; then
  # Create environment file with collected credentials
  # Add to .gitignore
fi

# Source the file
source .env.blueprint

4. Architecture Analysis

# Detect blueprint structure
if [ -d "network" ] && [ -d "clusters" ]; then
  architecture_type="multi-layer"
  layers=("network" "clusters" "nodes")
else
  architecture_type="single-layer"
  layers=(".")
fi

# For each layer, run terraform plan to count resources
cd ${layer_dir}
terraform init -backend=false
terraform plan -detailed-exitcode
# Parse plan output for resource count

5. Deployment Orchestration

For multi-layer deployments, use Task tool to spawn subagents:

Task 1: Deploy network layer
  Prompt: "Deploy the network layer at $blueprint/network.
          Run terraform init, plan, and apply.
          Capture outputs for next layer.
          Report progress with resource creation status."

Task 2: Deploy frontend cluster (after Task 1)
  Prompt: "Deploy the frontend cluster at $blueprint/clusters/frontend.
          Use data sources to read network layer state.
          Run terraform init, plan, and apply.
          Report cluster ARN and endpoint when complete."

Task 3: Deploy backend cluster (parallel with Task 2)
  [Similar prompt]

For single-layer deployments, execute directly.

6. Validation

# Verify Aviatrix resources
# Parse terraform output for gateway names

# Verify Kubernetes resources (if applicable)
aws eks update-kubeconfig --name <cluster-name> --region <region>
kubectl get nodes
kubectl get pods -A

# Run connectivity tests if defined in README

7. Post-Deployment Output

# Extract important outputs
terraform output -json > outputs.json

# Parse outputs and format access instructions
# Include:
# - Cluster names and access commands
# - Controller/CoPilot URLs
# - Cost summary
# - Next steps from README
# - Destroy instructions (in reverse order)

Error Handling

Terraform Errors

If terraform apply fails:

1. Capture full error output
2. Check common issues (missing variables, auth failures, quota exceeded, dependencies)
3. Provide specific remediation steps
4. Ask if user wants to retry, continue, or abort

Credential Issues

If authentication fails:

1. Verify credentials in .env.blueprint
2. Test Controller connectivity
3. Test cloud provider credentials
4. Offer to recreate .env.blueprint

Partial Deployment

If deployment fails mid-layer:

1. Identify which resources were created
2. Offer options: resume, destroy partial deployment, or manual intervention
3. Preserve state files
4. Log all outputs for debugging

Plan-Only Mode

When --plan-only flag is used:

1. Run all checks through Phase 5 (Planning)
2. Execute terraform plan (but not apply) for each layer
3. Show complete resource list and cost estimates
4. Export plan to file: terraform plan -out=tfplan
5. Provide command to apply later

Layer-Specific Deployment

When --layer <name> flag is used:

1. Verify dependencies are met (previous layers deployed)
2. Deploy only the specified layer
3. Skip layers that depend on this one
4. Warn about incomplete deployment

Output Files

Generate these files during deployment:

• .env.blueprint - Environment variables with credentials
• .deploy-blueprint-<timestamp>.log - Complete deployment log
• terraform.tfplan - Saved plan file (if --plan-only)
• outputs-<layer>.json - JSON outputs from each layer

Security Notes

• Never display passwords in plain text (use ******)
• Always add .env.blueprint to .gitignore
• Warn about local state file risks
• Remind users to destroy resources when done
• Check for exposed security groups or public endpoints in validation