[X-Guardian]

Description

Follow-on to chore(gha): optimise examples CI pipeline (fabb452). Applies the same patterns to integration.yml and provider-integration.yml, plus a couple of waste-reduction wins specific to those workflows.

Integration tests (integration.yml)

• Matrix shaped per-target instead of cartesian. tools/build-test-matrix.mjs now inspects each test file and emits hclOutput: true only when the file contains an HCL-positive decorator (onlyHcl / onPosixWithHcl / onWindowsWithHcl). Tests that only use onlyJson — or no decorator at all — previously paid a full beforeAll setup cost in HCL mode to skip every assertion. Matrix size drops from 240 → 148 jobs (~38% reduction); 60 unique test targets are preserved, coverage is identical.
• Removed the tools/align-version.sh step. The matrix consumes dist/ via Verdaccio (run-against-dist); the version string is irrelevant for non-publish runs.
• Terraform and go cache keys de-scoped: terraform-<os>-<tf>-matrix-integration-<target> → terraform-<os>-<tf>-integration. Stops 60 matrix jobs each writing their own cache entry.
• made dist artifact name unique.

Provider integration tests (provider-integration.yml)

• Same align-version removal and cache-key de-scoping.
• Matrix shaped per-template instead of cartesian. All 12 providers use test/provider-tests/template/test.ts, which today contains only an onlyJson block — so the 12 hclOutput: true jobs did zero useful work. The new tools/build-provider-test-matrix.mjs grep-checks the template for HCL-positive decorators and only emits HCL entries when meaningful. Matrix size drops from 24 → 12 jobs (50% reduction); will auto-double if the template gains an onlyHcl assertion.
• made dist artifact name unique.

Tooling cleanup

• tools/build-test-matrix.sh and tools/build-provider-test-matrix.sh ported to ESM (.mjs), matching the style of the existing tools/build-example-matrix.mjs.
• The disabled windows_integration block in integration.yml (gated if: false) was brought up to the new shape so a future re-enable doesn't regress.
• Added name: fields to previously unnamed steps across both workflows so the Actions UI surfaces each step clearly.

Impact

Comparison: PR 178 (SHA 4eeb83f, 2026-05-18) vs baseline PR 175 (SHA 6eb6ff0, 2026-05-15)

Integration Tests

	Baseline (PR 175)	PR 178	Delta
Matrix jobs	243	150	−38%
Aggregate compute (sum of job durations)	14.0h	7.4h	−47%
Suite wall-clock (min job-start → max job-end)	3h 23min	1h 35min	−53%

Provider Integration Tests

	Baseline (PR 175)	PR 178	Delta
Matrix jobs	50	25	−50%
Aggregate compute	2.5h	1.3h	−47%
Suite wall-clock	1h 10min	36min	−49%

Checklist

• I have updated the PR title to match CDKTN's style guide
• I have run the linter on my code locally
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation if applicable
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works if applicable
• New and existing unit tests pass locally with my changes

[github-actions]

I'm going to lock this pull request because it has been closed for 30 days. This helps our maintainers find and focus on the active issues. If you've found a problem that seems related to this change, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.