| name | testing-spec-kit-canon-extension |
| description | Validate the local Spec Kit Canon extension and canon-core preset against the real sibling workspace layout (`spec-kit`, `spec-kit-canon`, `spec-kit-canon-test`). Use when starting or resuming the dedicated extension test workflow, optionally clearing `spec-kit-canon-test` before step 1, overriding the scaffolded Spec Kit script variant (`sh` vs `ps`) when needed, verifying constitution config rendering from the shared fixture, seeding the trivial canon, running the standard `/speckit.specify` to `/speckit.implement` workflow, exercising `/speckit.canon.drift` (with manual resolve and analyze gates) and `/speckit.canon.vibecode-drift` (with analyze gate before canonize), verifying that canon updates capture new API and web UI behavior, and generating a final Markdown test report with per-step results, test-run metrics, and developer-history copies in `spec-kit-canon/tests/history/`. |
Testing Spec Kit Canon Extension
This folder is the canonical shared skill source.
Use the shared material from:
./core
./scripts
./assets
Shared Rules
- Treat core/workflow.md as the canonical workflow.
- Treat core/test-flow.md as the canonical prompt and
verification reference.
- Begin command-driven runs with
scripts/prepare_workflow.py so resume versus
restart behavior stays consistent across agents.
- Use scripts/manage_progress.py to create and
update the resumable workflow progress file.
- Use scripts/generate_test_report.py to
render the final Markdown report from the saved workflow state.
- Treat
--script sh as the default fresh-run variant unless a command
explicitly requests --script ps.
- Resume the current workflow by default when the progress file still has
pending work.
- If the progress file shows the workflow is fully completed, restart from
scratch automatically before continuing.
- Support a command-level
--restart flag that forces a fresh run even
when resumable progress exists.
- Keep the shared workflow and validation guidance in this skill package.
- Treat
spec-kit-canon-test as a real git repository. The validation run must
leave it in a valid branch/commit state, not just a passing working tree.
- Before staging sandbox changes, verify the installed extension copy at
.specify/extensions/canon does not contain a nested .git directory. If it
does, remove the installed copy's .git metadata before committing the
sandbox and note the packaging regression in the source repo.
- Before switching branches during the workflow, make sure the progress file at
.specify/tmp/testing-spec-kit-canon-extension-progress.json is committed or
stashed if git would otherwise block checkout.