jobtrackingapp/docs/redesign/phase-5/README.md

Phase 5 — Representative screens / major UI redesign plan

Current successor: P5-04b query readiness, GO for the shared-read prerequisite. P5-01–P5-04a implementation records remain authoritative for their scopes. The separate Table and Board now share proven logical query membership; Table/Board integration has not begun. P5-02 manual screen-reader acceptance remains mandatory for P5-15 and does not block independent Jobs work. P2-D01, F04, native touch verification and recorded baseline timing observations remain open. B07 has not started.

Historical planning checkpoint: source 2a2f9e48052c562ef599825458d0d3a6f9f9fc61, inspected on 2026-09-13 on codex/ui-redesign-phase4. That checkpoint was documentation-only; the successor record above owns implementation/validation status.

P4-06 gave a CONDITIONAL GO for planning. The original plan required the separate P5-00 gate before P5-01 AppShell/grouped navigation, the first major visible redesign. Use the successor's explicit acceptance conditions; implementation success is not a production release approval.

Package

• Implementation plan: 17 bounded checkpoints, dependencies, rollback and stop rules; P5-05 preview deliberately reserved/deferred.
• Readiness gates: exact shell, package, revision, rendering, query and release prerequisites.
• Navigation migration: every current menu/utility, all 31 existing patterns, staged future destinations and precise shell direction.
• Preservation impact: all 57 original contracts, current evidence, planned changes/gaps and visual trace IDs.
• Visual acceptance: stable states/viewports/data and manual/automated evidence requirements.
• Open blockers: inherited defects, slice deadlines, separate backend/product/production gates.

Source reconciliation and findings

Read alongside the original redesign sequence, Phase 1 complete inventory, Phase 2 matrix and trace format, Phase 3 architecture/impact, P4-01, P4-02, P4-03, P4-04, P4-05 and especially P4-06 acceptance / 57-row reconciliation. Historical future-tense status in older packages is not current implementation state.

Current source inspection confirms:

• App has 31 page/alias patterns, a session-invalidated provider-tree unmount, existing shell groups/collapse and distinct account/Career Profile routes. Preserve React Router and providers; table preferences still participate in router memoization, so characterize rather than assume state lifetime.
• At planning time JobTable and KanbanBoard did not share filter/read scope. The separate readiness checkpoint resolves that prerequisite while retaining complete Board population and paged Table responses. Integration still must not substitute a single Table page for the Board.
• QuickCommandDialog has six commands plus job/company results, not global CV/document search. Notifications already have their own store/actions. Reuse these and repair explicit error/keyboard states, without new producers.
• Analytics reports exist inside Dashboard with two load groups and five owner-local visibility flags. /analytics is a new destination for existing data, not permission to invent metrics. Documents/global aggregate Admin data are not ready; do not add empty navigation targets.
• The CV editor already has Template/Content/Customize/AI/History, live preview, overrides, undo/history, mobile panes and extensive controls. Keep them. Server revision enforcement remains absent on Save, and preview injects pagination/cropping into rendered HTML. Those facts justify independent integrity/rendering gates before broad CV changes.
• A production quick-preview host is absent. The optional overlay props/test are not proof of an existing launched feature. NEW-10 stays outside the default implementation queue.

No contradiction requires a new component library or platform. P4 shared MUI semantics, focus, PageHeader/PageSection and opt-in density remain the foundation. Career Profile/library/complete Documents, Settings/Admin/mail and full Notifications remain Phase 7; advanced dashboard layout remains Phase 8. The current user request/Phase 3 sequencing advances only the bounded existing-report Analytics presentation into Phase 5.

Validation baseline

The following is inherited executed evidence from P4-06 at the source checkpoint, not newly executed application tests during this documentation-only task:

Gate	Unchanged baseline
Backend Release tests	916 passed
Frontend Jest	354 passed, 75 suites
Chromium browser	51 actual passes + unchanged P2-D01 expected failure (reporter 52 includes the expected failure)
Rate-limit / ingress / PDF admission	9 / 6 / 1 passed
Configuration/Compose	3 passed
TypeScript / lint / frontend production build / API Release build	Passing; API zero warnings/errors; P4-06 also checked edited e2e TypeScript
Secret/whitespace/docs checks	Passing at baseline; fresh documentation checks below

Application/test/CI counts change by zero here. No test or CI configuration is edited. Future implementation checkpoints run the exact complete commands recorded in P4-06 final validation, plus new scoped tests and relevant e2e TypeScript. Specialized browser suites own the same disposable API port and run sequentially. Preserve B01–B06 checks; no production evidence follows from them. After a separately accepted P2-D01 repair, report the transition from expected failure to actual successful native archive proof explicitly, without renaming or masking it.

P4-06's 118 retained axe scans include eight known rule occurrences and many incomplete checks. Mapping 57 contracts is not exhaustive behavior, visual, provider or accessibility coverage. This plan neither changes those results nor treats the incomplete nodes as passes.

Planning checkpoint verification

Fresh local planning checks (2026-09-13):

Executed command / review	Result
python .artifacts/p5-check-plan.py	Seven required documents; 57 unique PRES IDs exactly match Phase 2/P4-06; 31 unique current routes exactly match App; 17 planned implementation checkpoints; 184 local link/anchor checks and 127 reference-style test-file targets resolve; table columns/fences pass
python .artifacts/p4-04-check-doc-links.py	131 current/root Markdown documents, zero missing local targets; historical archive excluded
Source review of App/AppShell, command/notification/table/board/dashboard/workspace/CV owners and service reads	Reconciled actual route/provider lifetime, absent preview host, distinct board population, existing analytics and CV revision/preview boundaries; no runtime/provider result inferred from source
git diff --exit-code 2a2f9e48052c562ef599825458d0d3a6f9f9fc61 -- job-tracker-ui JobTrackerApi JobTrackerApi.Tests scripts .gitea browser-extension	No application/backend/tests/build/CI/infrastructure changes
git diff --cached --name-only / git diff --cached --stat	Eleven Markdown files only: seven package files, three index/trace links and TODO
python scripts/supply-chain.py scan-secrets	Staged documentation secret check passed
git diff --check / git diff --cached --check	Passed; ordinary repository LF/CRLF conversion notices are not whitespace failures

The two .artifacts validators are ignored local review tools, not new CI/application files. They validate documentation structure/targets, not future implementation behavior. Full application suites were not rerun for prose; their last actual evidence is the unchanged baseline above. Git push is limited to the requested feature branch; remote SHA and clean-tree status are verified and reported with the final commit. No production deployment or external provider check is claimed.

Recommended next instruction

“Implement P5-00 — Shell readiness gates only from the documentation checkpoint, following READINESS_GATES.md and PHASE5_PLAN.md. Reproduce/protect each prerequisite against the current UI, preserve application behavior, run the full baseline, record real visual/accessibility evidence, commit/push the bounded readiness checkpoint, and stop before P5-01.”

P5-01 begins the first major visible redesign only after P5-00 acceptance. Native package proof blocks dependent section/relationship migration; revision/rendering proof blocks broad CV; query parity blocks Board unification. Independent screens need not wait for those gates. F04 remains open for production verification and does not block local UI planning. A feature-branch commit/push is not permission to deploy.