메뉴
Deploy an Aviatrix blueprint to your environment with guided, multi-step orchestration. Handles both simple single-layer and complex multi-layer blueprints.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Tear down an Aviatrix blueprint and clean up all cloud resources. Handles both simple single-layer and complex multi-layer blueprints in the correct reverse order.
AI-QA an Aviatrix blueprint by deploying it as a customer using only the README, capturing UX gaps, and opening a fix PR. Replaces test-blueprint.
Pre-QA quality gate for Aviatrix blueprints. Runs tiered validation (blockers / standards / quality) and produces a QA-readiness verdict. Use before submitting a blueprint to QA, before merging, or whenever asked to "validate", "check", or "grade" a blueprint.
Analyze a blueprint directory and generate comprehensive documentation about what it creates and requires. Use when the user asks to analyze, understand, or document a 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 an Aviatrix blueprint to your environment with guided, multi-step orchestration.
/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.
This skill requires:
If no blueprint path provided, display available blueprints and prompt for selection.
Display key sections from the blueprint README:
Ask for confirmation to continue.
Verify all required tools and access:
Interactively prompt for:
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
Analyze blueprint architecture:
terraform plan for each layer to count resourcesAsk for confirmation to proceed.
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.
Verify deployment success:
Provide access information:
When this skill is invoked:
# 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
# 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:
# 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
# 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
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.
# 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
# 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)
If terraform apply fails:
If authentication fails:
If deployment fails mid-layer:
When --plan-only flag is used:
terraform plan -out=tfplanWhen --layer <name> flag is used:
Generate these files during deployment:
.env.blueprint - Environment variables with credentials.deploy-blueprint-<timestamp>.log - Complete deployment logterraform.tfplan - Saved plan file (if --plan-only)outputs-<layer>.json - JSON outputs from each layer