Skip to content

quality.md

Latest commit

 

History

History
63 lines (53 loc) · 4.45 KB

quality.md

File metadata and controls

63 lines (53 loc) · 4.45 KB
Raw

Quality dashboard

This page brings the repository's quality evidence into one place. A green badge means the latest main run of that workflow passed. Some specialized workflows are path-filtered, so a green badge is evidence about the files and commit that triggered it, not a blanket statement that every part of the repository was checked. The main CI workflow is unfiltered and runs its Python and site matrices on every pull request. For a pull request, use that pull request's checks rather than these default-branch badges.

Live signals

Signal Status What it establishes
Reporting site end to end e2e status Playwright builds the reporting site and checks the rendered dashboard against the committed run data.
Repository CI CI status Python 3.12/3.13 run every pytest contract and enforce production Python coverage, including Kubernetes 1.36 manifest schemas, Argo wiring, workload resource bounds, and the deny-by-default agent-governance policy; Node.js 22 runs the reporting-site unit suite.
Reporting site deployment pages status The site unit tests and Astro build passed before the current GitHub Pages deployment.
Advisory AI review Claude review status For eligible same-repository pull requests, Claude checked the diff against the repository review and security guidance; its comments still require human verification.

The operational QA dashboard at https://frostyard.github.io/lab/ is a different signal: it reports the images and QA lanes exercised by Argo. It does not report whether this repository's code passed the checks above.

Evidence expected by change

Changed area Expected evidence
site/, e2e/, or Playwright configuration just site-e2e; run cd site && npm test when changing the API/data helpers.
scripts/, tests/, or policies/ python -m pytest -q; for governance changes also run python3 policies/check_agent_governance.py.
argo/, manifests/, argocd/, or Kubernetes resources python -m pytest -q for offline schema and cross-resource contracts; also run argo lint where applicable and just validate against a configured cluster.
Documentation Commands and links resolve, and claims about lane status agree with README.md and docs/roadmap.md.
Every pull request Follow the contributing guide and record the relevant result in the pull request's Testing section. Same-repository, non-draft pull requests also receive the advisory Claude review.

Known gaps

  • The E2E workflow runs on pull requests only when its path filters match. The unfiltered CI workflow runs repository unit, policy, and offline manifest tests on every pull request.
  • CI rejects malformed or duplicate-key YAML, strictly validates built-in resources against Kubernetes 1.36 schemas, and checks Argo references, Argo CD ownership, workload CPU/memory bounds, RBAC, and image-poller state. It cannot execute Argo CRD admission or cluster-specific policy without cluster credentials, so argo lint and just validate remain additional review evidence for manifest changes.
  • The Pages workflow runs after changes reach main; it is deployment evidence, not a pre-merge check.
  • Claude review requires the ANTHROPIC_API_KEY repository secret and skips fork pull requests. It is advisory evidence, not approval or a required replacement for deterministic checks and maintainer review.

These gaps must stay visible until a workflow actually closes them. Adding a new check should update both the live-signals table and the change-area mapping above.