| name | terraform-destroy |
| description | Destroys Terraform stacks by running terraform destroy. Supports destroying a single
stack by folder name or all stacks in reverse numeric order (highest number first, so
dependent stacks are torn down before their dependencies). The remote backend stack
(00-remote-backend-stack) is never destroyed — it is always excluded automatically.
Asks for confirmation exactly once before starting — after that, all stacks are
destroyed without further prompts. Use this skill whenever the user wants to destroy,
tear down, remove, or clean up Terraform infrastructure in this project. Also trigger
when the user says things like "destroy all stacks", "tear down the VPC", "tf destroy",
"nuke the infra", "clean up everything", or "remove the network stack".
|
Terraform Destroy
Destroy Terraform stacks in reverse dependency order with a single confirmation.
Protected Stack
The 00-remote-backend-stack is never destroyed. It holds the S3 bucket that stores
Terraform state for all other stacks — destroying it would orphan every other stack's
state. This exclusion applies whether destroying all stacks or targeting it by name.
If the user explicitly asks to destroy the remote backend stack, explain why it's
protected and refuse.
Arguments
This skill accepts an optional stack name argument:
/terraform-destroy — destroys all stacks (except the remote backend) in reverse numeric order/terraform-destroy <stack-folder> — destroys a single stack (e.g., /terraform-destroy 01-network-stack)
The stack folder name corresponds to a directory under terraform/ in the project root.
Pipeline
Step 1: Discover stacks
- If a specific stack was provided, verify the directory exists at
terraform/<stack-folder>/. If it is 00-remote-backend-stack, refuse and explain why.
- If no stack was provided ("destroy all"), list all directories under
terraform/, exclude 00-remote-backend-stack, and sort the remaining stacks in reverse numeric order. Higher-numbered stacks are destroyed first because they depend on lower-numbered ones.
Step 2: Confirm once
List all stacks that will be destroyed and ask for confirmation exactly once:
The following stacks will be destroyed (in this order):
1. 01-network-stack
Note: 00-remote-backend-stack is protected and will NOT be destroyed.
Type "yes" to confirm.
If the user confirms, proceed with all stacks without asking again. If they decline, stop.
Step 3: For each stack, run destroy
For each stack directory, in order:
3a: terraform init
Ensure the stack is initialized. Check if backend.hcl exists — if so, use terraform init -backend-config=backend.hcl. If .terraform/ already exists, skip init.
3b: terraform destroy -auto-approve
Run terraform destroy -auto-approve — no per-stack confirmation since the user already confirmed in Step 2.
- Show the destruction summary (resources destroyed count)
- If destroy fails, show the error and ask if the user wants to continue with the remaining stacks or stop
Step 4: Summary
After all stacks have been processed, provide a brief summary:
- Which stacks were destroyed successfully (and how many resources each)
- Which stacks failed (and the error)
- Which stacks were skipped
- Remind that
00-remote-backend-stack was preserved
Important Notes
- Always
cd into the stack directory before running terraform commands
- Reverse numeric order is critical — higher stacks depend on lower ones
- The single upfront confirmation replaces per-stack confirmations — once the user says yes, execute everything
- Environment variables and AWS credentials should already be configured in the shell
