terraform-test skill: incorrect examples and CLI guidance (plan-time unknowns, variable precedence, -filter, undefined cross-run output)

[jrx]

Found while vendoring the skill into a downstream repo and running an automated review over the vendored files. Four issues in terraform/code-generation/skills/terraform-test/:

1. Plan-mode example asserts an apply-time value (SKILL.md)

The basic test file example uses command = plan with:

assert {
  condition     = aws_instance.example.id != ""
  error_message = "Instance should have a valid ID"
}

id is provider-computed and unknown at plan time, so the condition cannot evaluate to a known boolean and the run fails instead of demonstrating a working pattern. The example should assert a plan-known attribute such as instance_type, or switch to command = apply.

2. Contradictory variable precedence comment (SKILL.md)

The test file structure example says:

# Optional: file-level variables (highest precedence, override all other sources)

but the same skill (and Terraform itself) documents run-level variables blocks as overriding file-level ones. File-level values are defaults for runs, not the highest-precedence source.

3. -filter comment is wrong (SKILL.md, Running Tests section)

terraform test -filter=test_vpc_configuration        # by run block name

-filter selects test files (e.g. -filter=tests/defaults.tftest.hcl), not run blocks. There is no flag to run a single run block.

4. Cross-run reference to an undefined output (references/EXAMPLES.md)

run "test_subnet_references_vpc_with_mocks" uses run.setup_vpc_with_mocks.vpc_id, but setup_vpc_with_mocks defines no vpc_id output, so the example cannot execute. Either add output "vpc_id" { value = aws_vpc.main.id } to the module under test and reference it, or drop the cross-run reference.

Happy to send a PR for any or all of these if useful.