- Select the hand-authored
site/files and public SVGs as the canonical source. - Record the GitHub Pages topology in ADR-010 and define Website requirements and acceptance evidence.
- Add a failing repository contract test for source triggers, immutable Pages actions, deployment permissions, the custom domain, and public assets.
- Add the GitHub Pages workflow around the parallel site, then pass the focused test.
- Validate the staged static artifact and rendered desktop/mobile site.
- Commit only Website delivery files plus the explicitly identified site
content, push
main, and observe the real Pages deployment to completion. - Configure and verify the Pages custom domain, HTTPS, GitHub About URL, repository topics, and live canonical page.
Completion requires the remote workflow to finish successfully and the custom HTTPS URL to serve the published revision. A local render or pushed commit is not delivery evidence.
- Inspect KeyLoad's JSON artifact, aggregation and Pages publication flow.
- Extend ADR-010 with schemas, independent provenance and the artifact trust boundary; register REQ/AC/TASK/TEST-WEB-002.
- Add failing real-process regressions for TRX-to-JSON, benchmark JSON, invalid/partial evidence and immutable publication.
- Export per-runner test JSON and aggregated numerical benchmark JSON.
- Collect trusted completed
mainartifacts and build publicdata/latest.json; trigger Pages after either source workflow finishes. - Render current test counts and scoped/filterable benchmark metrics, provenance, nullable values and missing evidence on desktop/mobile.
- Run format/build/analyzers, focused regressions, security and
architecture checks, and focused coverage; record exact evidence.
Local 19/19 Website, 2/2 hosted compatibility and 32 Rust tests pass;
Website source coverage is 92.64%. See
docs/Development/WebsiteMetricsEvidence-2026-10-03.md. - Commit scoped changes, push and observe JSON publication and Pages deployment; verify the canonical HTTPS metrics endpoint.
Keep the task in_progress until remote publication and live evidence exist.
This does not close the statistical benchmark or broader runtime plans.