Sample implementation demonstrating the iac-spec-kit workflow for infrastructure as code development.

Reference: WordPress on IBM Cloud workflow

Input: "This is a small business production website. Keep it simple but include security and automated backups. Keep costs reasonable."

Output: principles.md

Input: "I need to deploy WordPress for my small business website. Should handle a few thousand visitors per day, needs to be secure with automated backups. Budget is around $500/month."

Output:

• spec.md
• checklists/requirements.md

Input: Answered 5 clarification questions (EU region, ephemeral containers, user-managed DNS, auto-expand storage, alert on budget)

Output: Updated spec.md

Output:

• plan.md
• research.md
• architecture.md
• modules.md
• quickstart.md

Output: tasks.md - 110 tasks across 8 phases

Output: 17 Terraform files in iac/

Validation: terraform init ✅ terraform validate ✅ terraform plan ✅ (62 resources)

This implementation required iterative refinement beyond initial code generation. The aspects below are specific to the AI model used, and the AI tool used. The point here is that the framework provides a SSD driven methodology to catch issues early, before code generation, but as with any AI-driven flow, requires human expertise in the loop, for architectural decisions and error interpretation.

Architecture correction: Initial design included VPC infrastructure. During implementation, I pointed out that Code Engine is a managed service using Cloud Service Endpoints, eliminating need for VPC/VPE gateways.

Validation-driven fixes: While terraform validate passed, terraform plan revealed a few configuration errors that were fixed by inputing the errors in the AI tool.