Repository navigation
Sprint 2
October 14 through October 27, 2026. Owner: Matthew Feroz.
Demonstrate that a new evaluator can use the released workflow with a compatible installed worker, then leave repeatable evidence and a clear maintainer release procedure. Complete the remaining first-run, report-access and reproducibility acceptance from Sprint 1. Current public source already contains their implementation fixes; this sprint verifies the user outcomes rather than rebuilding the features.
Detailed requirements and observable completion criteria are in the assigned issues.
| Issue | Work | Owner | Estimate |
|---|---|---|---|
| #16 | Validate first-run setup with a fresh Clerk account | @MatthewFeroz | 6 h |
| #17 | Verify hosted report publication and access removal in real browsers | @MatthewFeroz | 6 h |
| #18 | Publish a reproducible Oracle example and maintainer release checklist | @MatthewFeroz | 6 h |
| #19 | Verify source-built worker upgrades against Harbor 0.24 | @MatthewFeroz | 6 h |
| #20 | Refresh public prototypes and architecture for the released workflow | @MatthewFeroz | 4 h |
Milestone · Issue list · Project
The commitment is 28 estimated hours plus 4 hours of contingency, matching the previous two-week planning capacity. Keep one issue actively in progress at a time. Estimates are planning judgments, not actual time logged.
- Week 1: fresh real-account Oracle setup and the complete report-access browser path. Begin isolated installed-worker compatibility checks.
- Week 2: finish upgrade/rollback evidence, publish the reproducible Oracle example and release checklist, then reconcile public prototypes/reference pages with the verified flow.
The original Sprint 1 issues #9, #11 and #12 retain their milestone and history. New issues #16, #17 and #18 track only their remaining work. Close the parent issue only after its remaining completion criteria pass. #10 is complete with linked recovery evidence. Neither old status report is rewritten.
Each completed issue must link a tested revision, appropriate check results, sanitized evidence and remaining limitations. Use Oracle and clearly labeled synthetic reports; no paid model calls are needed. Check actual frontend/backend, package and Harbor versions before making deployment claims. Installed CLI validation uses Node and the exact artifact under review.
Use a dedicated worker and isolated account/data state. Deployment, npm publication, shared-worker changes, a new collaboration system and paid studies are outside this sprint commitment and require separate authorization. A reported failure receives a focused fix; unfinished criteria remain open rather than being marked Done.