Skip to content

chore(openspec): archive add-ci-and-npm-publish-automation - #39

Merged
kewang merged 1 commit into
masterfrom
chore/archive-ci-and-npm-publish
Aug 5, 2026
Merged

chore(openspec): archive add-ci-and-npm-publish-automation#39
kewang merged 1 commit into
masterfrom
chore/archive-ci-and-npm-publish

Conversation

@kewang

@kewang kewang commented Aug 5, 2026

Copy link
Copy Markdown
Collaborator

Archives the change behind #36. The CLI's automatic sync succeeded — no manual delta application needed, and openspec validate --specs passes 50/50 (up from 48).

Two new capabilities, both purely additive:

Capability Requirements Scenarios
continuous-integration 9 22
npm-package-cicd 5 14

Task 6.8 removed rather than left unchecked. Reaching the OIDC authentication step needs a real npm publish, and neither package's src/ had changed since its last release — so under this change's own rules neither was due for one, and publishing an identical tarball under a new number just to exercise the pipeline is a bad trade against a registry that cannot be un-published.

design.md now carries both a Verified at implementation time section and a Knowingly unverified one. What ran for real: the version comparison against the live registry, the check job on a real runner, job-to-job output passing, and the skip path (exercised when #36 merged — both packages skipped, nothing published, run green). What has never executed: npm publish itself. Finding out late is cheap — a failed publish leaves the registry untouched, and the tag is created only after a successful one.

Touches openspec/ only.

Automatic sync succeeded — 14 requirements written into two new main specs,
no manual delta application needed. openspec validate --specs: 50/50.

- continuous-integration: 9 requirements, 22 scenarios
- npm-package-cicd: 5 requirements, 14 scenarios

Task 6.8 (end-to-end publish verification) was removed rather than left
unchecked. Reaching the OIDC authentication step requires an actual npm publish,
and neither package's src/ had changed since its last release, so under this
change's own rules neither was due for one — publishing an identical tarball
under a new number purely to exercise the pipeline is not a trade worth making
against a registry that cannot be un-published. design.md records what is
verified (version comparison, check job on a real runner, job-to-job outputs,
the skip path running for real on master) and what is not, plus why finding out
late is cheap: a failed publish leaves the registry untouched, and the tag is
created only after a successful one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PgfFLCcVjPxZJmMspbY4Bu
@kewang
kewang merged commit 7c412e5 into master Aug 5, 2026
3 checks passed
@kewang
kewang deleted the chore/archive-ci-and-npm-publish branch August 5, 2026 02:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant