How far would you trust generated Terraform before reviewing it manually?
I’ve been working on Terraform generation for multi-cloud architectures and I keep coming back to the same problem:
Generating valid HCL is the easy part.
Generating infrastructure I would actually trust is much harder.
Right now I’m looking at the generated stack in several layers:
provider-specific Terraform resources
variables and outputs
remote state configuration
networking and IAM
security validation
cost estimation
documentation explaining the architecture
The difficult part isn’t generating those files. It’s deciding what should be automatically generated versus what should require an engineer to explicitly approve it.
For example, I wouldn’t want a generator silently making important IAM, networking or state decisions just because the Terraform validates successfully.
So for people running Terraform in production:
What would you need to see before you’d trust generated Terraform enough to actually run terraform plan on it?
Tests? Checkov/tfsec? Policy-as-code? A generated architecture diagram? Cost diff? Something else?