BaseballHelm finish: #379 stats reconciliation, heavy mobile pass, CI gates, tests — 45-PR overnight batch - #856
Merged
Merged
Conversation
The app write path (createCamp/updateCamp in src/app/baseball/actions/camps.ts) always writes status:'published' because the baseball_camps_select RLS policy only exposes non-owner (player) reads when status = 'published'. The E2E seed fixture still inserted the seeded "E2E Prospect Camp" with status:'active', making it invisible to the player-facing camps browse query and silently failing the "Camps - Player Flow" specs in e2e/camps.spec.ts. No schema/migration change needed — baseball_camps.status is a free text column with no CHECK constraint, and 'published' is already the value the real app writes. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…gREST cap (#813) getEliteStatEvents built its 4 event-grain queries (baseball_pitch_events, baseball_batted_ball_events, baseball_swing_events, baseball_plate_appearances) with no .order()/.range() at all, so any team with more than 1000 rows in a table (a single season of pitch-by-pitch data easily exceeds this) silently had derived metrics — CSW%, chase/whiff rate, velo-decay, pitch mix, hard-hit rate, command heatmap — computed from an arbitrary 1000-row slice, not even a stable "most recent 1000". Convert all 4 buildQuery(...) call sites to fetchAllRowsResult with a stable .order('id', {ascending:true}) + .range(from,to), mirroring the identical, already-proven pattern coachhelm/engine-run.ts uses against these same tables. The existing team/staff-gate envelope and error-branch shape are unchanged; only the data queries are now paginated. Adds a focused pagination regression test (the file's own existing test suite is deliberately pure-math-only per its header) that simulates the real PostgREST 1000-row-per-response cap in a mock query builder and asserts getEliteStatEvents returns all 1500 rows across 2 pages, issues a stable .order('id') before ranging, and still surfaces an honest error (no throw) on a failed page read. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…) (#814) PrimaryCtaRow rendered up to three peer pill buttons (Check In, Acknowledge, View Today Plan/Schedule) with no priority collapse. Now resolves a single filled primary CTA by priority (pending ack > due check-in > view plan/schedule per the issue's acceptance criteria) and demotes any remaining applicable action(s) to inline text links — visible, not hidden, just secondary. Also reorders NextEventHero ahead of the one-time "Activate Recruiting" nudge card so real schedule content reaches the first viewport before a promotional card, addressing the issue's third acceptance criterion without touching unrelated component internals. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…port + scout-packet (#815) player-passport.ts and scout-packet.ts still summed innings-pitched with naive base-10 `+=` and divided ERA/WHIP by that raw notation sum, so 6.1 + 6.2 yielded 12.3 instead of the correct 13.0 — the exact bug #434 already fixed in stats-center.ts and BoxScoreView.tsx via the shared src/lib/baseball/innings.ts helper (sumInningsPitched/ipToInnings), but never wired into these two read models. Extracted the season pitching-line computation into a small pure, exported `summarizePitchingSeason(games, ipRows, totals)` in each file (same shape as stats-center.ts's addPitching/finalizePitching split) so the fix is directly unit-testable without mocking Supabase. Both files now sum IP via outs and derive era/whip from ipToInnings(ip) (true innings), never the notation value — mirroring the proven pattern exactly, no shape/signature changes to either read model's public output. Added src/lib/baseball/read-models/__tests__/{player-passport,scout-packet}-innings.test.ts proving the 5.2 + 3.2 -> 9.1 thirds case (not 8.4), the classic 6.1 + 6.2 -> 13.0 case (not 12.3), and that era/whip divide by true innings (28/3) rather than the 9.1 notation value. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…offsets (#485) (#816) CalendarFairway's outer shell derived its full-bleed height from a flat, hand-guessed `h-[calc(100vh-5.5rem-...)]` that predated the current AppShell (sticky glass top bar + a coach-only Team-hub sub-nav strip) and never matched either, producing double-scroll/dead-space on mobile. Replace it with two literal shell strings (coach vs. player) that subtract the shell's REAL offsets: the shared `--golf-mobile-header-offset` CSS var (top bar), the Team-hub HubSubNav row (45px, coach-only — Calendar has no sub-nav for players, confirmed via resolve-active-hub.ts's player branch), and AppShell's own mirrored bottom-nav clearance — using 100dvh instead of 100vh so iOS Safari's dynamic toolbar can't leave a sliver. Kept the determinate-height ancestor PremiumCalendarClient's internal h-full/flex-1 scroll region needs (verified it collapses to 0 without one) rather than going fully in-flow like golf's native FairwayCalendar, since baseball's calendar still reuses that shared component verbatim. Added CalendarFairway.shell-height.test.tsx pinning the corrected class strings so the flat 100vh guess can't silently return. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…allAction (#394) (#819) All 10 exports across discover.ts (4) and insights.ts (6) ran only through the guard-free withAdminObserved observability decorator — no auth/context/ capability enforcement, no demo-write guard, no Sentry scoping. - discover.ts: wrap all 4 read-only exports (getDiscoverPlayers, getDiscoverTeams, getWatchlistIds, getStateCounts) with withBaseballAction (requireActiveContext:false, demoSafe:true, no requiredCapability — Discover is deliberately cross-team by product design). Impl bodies unchanged. - insights.ts: deleted generateTeamInsights/getTeamInsights as dead code — zero callers repo-wide, superseded by engine-run.ts's shared reconciliation core, and generateTeamInsightsImpl carried a live #472-class coachId/user.id domain-confusion bug that was never exercised because nothing called it. - insights.ts: migrated dismissInsight/markInsightAddressed/ submitInsightFeedback onto withBaseballAction (ownership-gated in-body, no requiredCapability, requireActiveContext:false — mirrors dev-plans.ts's precedent). Each export is now a thin try/catch wrapper around the guarded action so a thrown BaseballUnauthorizedError/BaseballDemoReadOnlyError/ BaseballActionError is translated back into the existing {success:false,error} envelope — PlayerInsightsPanel.tsx calls these inside an un-try/catch'd startTransition and depends on always getting a resolved value. - Relocated resolveCallerCoachId out of the 'use server' insights.ts into a plain module (src/lib/baseball/insights/resolve-coach-id.ts) so it stops being an accidental public server-action endpoint. - Updated feature-registry.ts's baseball_insights action-name manifest to drop the deleted/relocated names. - Added the @sentry/nextjs + @/lib/demo/baseball-config.server mocks discover-privacy.test.ts and insight-lifecycle.test.ts now need (the real withBaseballAction wrapper touches both); updated insight-lifecycle.test.ts's resolveCallerCoachId import path. Known accepted risk (per spec): withBaseballAction's AUTH step throws unconditionally on no session, vs. discover.ts's prior graceful empty-shape-on-no-session return. DiscoverClient.tsx already try/catches every call site and falls back to a visible error state, so this is a visible-but-handled edge case on an already-authenticated dashboard route, not a crash. Gates: typecheck clean; eslint --max-warnings 0 clean on all 6 touched files; targeted vitest (discover-privacy, insight-lifecycle, coverage-contract.observability) 60/60 passing across all 4 vitest projects. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…e-menu-button (#820) Batch-H cleanup, re-verified zero importers before deleting: - PlayerPassportCard.tsx + its index.ts barrel export: only self-references and stale "legacy" doc-comment mentions remained; PlayerPassportFairway replaced it everywhere. Deleted both. - header.tsx: repo-wide grep (import path + <Header usage) confirmed zero real importers left (the only <Header> JSX hits are an unrelated golf FairwayDrivingSpray component). isRedesignEnabled()'s always-true flag had made its legacy-chrome branch permanently dead. Deleted the whole file rather than collapsing the fork, per fleet-verifier guidance. - mobile-menu-button.tsx: header.tsx was its only importer (page-header.tsx's MobileMenuButton resolves to the unrelated components/golf/MobileMenuButton). Now orphaned, deleted. - mode-toggle.tsx: kept — it has a real importer (baseball/coach/ModeToggle.tsx, JUCOModeToggle), so it does not qualify as a dead sibling here even though that importer's own only caller was the now-deleted header.tsx. Left as a follow-up note rather than expanding scope. - flag.ts: corrected the now-stale "NOT dead: header.tsx" caveat on isRedesignEnabled() — the function is still very much alive via dozens of live src/components/fairway/** call sites, just no longer via header.tsx. - MessagesFairway.tsx: left untouched per task ownership (another task owns its stale doc-comment fix). Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…bels (#821) Chunk 1 (dead query): the public player profile's recruiting-interests select listed a nonexistent `school_name` column, which PostgREST rejects with 42703. The `{ error }` was never checked, so `recruitingInterests` silently resolved to `[]` and the Dream Schools section (gated on `recruitingInterests.length > 0`) never rendered. Drop the bad column from the select (the Row shape is id/interest_level/organization via the `organizations` join) and log the error so a future regression is visible. Chunk 2 (label/enum dedup): the 5 pipeline stage ids were hand-copied in 4 places that had already drifted apart in production — `getPipelineStageLabel` (utils.ts) called the `watchlist` stage "Prospects" while `PIPELINE_STAGES` (stages.ts) calls the same stage "Watchlist", both live on the same Pipeline board. Make `PIPELINE_STAGES` the single source: `getPipelineStageLabel` is now a lookup into it, and `WatchlistSchemas.updateStatus` derives its zod enum from `PIPELINE_STAGES.map(s => s.id)` instead of a separate literal array. Resolves the collision in favor of stages.ts's "Watchlist" (no code comment defended "Prospects"). Found but out of scope: `PipelineStatusDot` in the frozen src/components/ui/status-dot.tsx carries a 5th copy of the same label map (also says "Prospects" for `watchlist`) but appears unused anywhere in the app; left untouched per the frozen-file rule. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…xpense identity + denial paths) (#823) src/app/baseball/actions/travel.ts had zero test coverage (BASEBALLHELM_FEATURE_READINESS_MATRIX.md:73). Adds src/app/baseball/actions/__tests__/travel.test.ts pinning: - createItinerary writes created_by = ctx.activeCoachId; addExpense writes created_by = ctx.user.id — deliberately different identity spaces, asserted with distinct fixture values so a future mocking mistake can't silently collapse them. - the two-layer team-capability guard: withBaseballAction's own pre-check only validates ctx.activeTeamId; createItinerary/addExpense both carry an explicit re-check when the caller-supplied teamId differs, which this locks in as a real gate (not just the wrapper). - not-found + capability-denial paths for updateItinerary/deleteItinerary/deleteExpense, which resolve their team from the EXISTING row rather than caller input. - a discovered quirk: invalid create-itinerary input degrades to the generic BaseballActionError message, not the raw zod message, because schema.parse() runs inside the withBaseballAction-wrapped body and ZodError isn't in the wrapper's rethrow whitelist — the ZodError branch in createItinerary's own try/catch is presently unreachable. Test pins actual behavior; no production code changed. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…gated writes (#824) Adds src/app/baseball/actions/__tests__/documents-write-capability.test.ts, a new file exercising what documents-capability.test.ts (one assertion) and documents.test.ts (read-path uploader-embed regression only) leave untouched: the six write actions in src/app/baseball/actions/documents.ts. Seeds a real baseball_team_coach_staff row so requireBaseballCapability's actual resolution logic runs (not stubbed), mirroring dev-plans-coach-gating.test.ts's idiom. Covers: - capability-less rejection / capability-holder success on all six writes (upload/create/update/delete/uploadNewVersion/revert) - team-id forging is blocked: upload/create re-check capability against an explicit teamId/team_id that differs from the active team; update/delete resolve the gated team from the EXISTING document row, never the caller's active team; uploadNewVersion's optional teamId argument is proven to never feed the capability gate at all (only document.team_id does — the arg only shapes the storage path) - signed-URL call sites (getPreviewUrl, getVersionHistory, getTeamDocuments/ withFreshDocumentUrls, upload, uploadNewVersion, revertToVersion) assert the exact storage_path + 3600s TTL, including that revertToVersion re-signs the ORIGINAL reverted-to path rather than uploading a new file - deleteBaseballDocument's storage.remove is scoped to exactly one document's version paths, never a sibling document's Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…#825) savePractice/publishPractice/recordPracticeAttendance/getPlayerPractices and saveObjective/savePracticeRecap/runPracticeEffectiveness/setReviewDisposition had zero action-level tests (BASEBALLHELM_FEATURE_READINESS_MATRIX.md:65-66,142). Adds three new test files locking down the stage-then-swap block replacement, server-side publish validation gate, publish-only calendar attach + per-player notify, the attendance upsert's no-delete contract, player-view block visibility + anchor-dependent class-conflict detection, target_metric registry validation, recap reps clamping, the effectiveness engine's disposition- preserving upsert + fail-closed read guard + hardcoded staff_only visibility, and the pure practice-validation issue codes (no direct test existed before). Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…y, Signal Inbox, Video Library (#826) Extends src/contracts/baseball/ (unit-project lane) with 6 new pinning test files covering the 3 surfaces #377 named but the existing contract suite never exercised directly (every prior reference to getPlayerToday/ getSignalInbox mocked the read model itself instead of running it): - product-trust/player-today-honest-loop.test.ts — getPlayerToday never fabricates a green readiness band (no check-in -> available:false; a submitted illness check-in surfaces the real red band/reasons via the UNMOCKED computeReadiness); assignments/coachActions/tasks/coachNotes degrade a sub-read failure to available:true+items:[]+a distinguishing `error` string, never conflated with genuine emptiness; a hand-entered recent-stat row never gets a fabricated trust/provenance object. - access/player-today-self-scope.test.ts — self-only authorization envelope (no baseball_players row, or membership on a different team -> authorized:false, never a caller-suppliable player id); coachNotes scope isolation (staff_only, archived, and a different player's note never leak into the player-visible feed). - coachhelm/signal-inbox-evidence.test.ts — getSignalInbox's OWN honesty invariants (distinct from signal-evidence.test.ts's signalFromInsight promotion-gate pin): sampleTooSmall catches a sub-threshold sample_n even when a generator left disposition at 'new'; confidence is never fabricated (null stays null, a legacy 0-100 value is honestly normalized); sourceRefs reflect real persisted refs; sub-read failures are distinguishable from genuine emptiness. - access/signal-inbox-staff-scope.test.ts — staff-only envelope + cross-team signal isolation in the same table. - product-trust/video-honest-empty-state.test.ts — Event/Tagged/Evidence views' `hasVideoEvents` gate distinguishes "no film at all" from "film exists but none qualifies for this view." - access/video-visibility-scope.test.ts — Library/Player views scoped through the requested team's roster; Event/Tagged/Evidence scoped by team_id; Evidence's signal-metadata enrichment never leaks a different team's signal title even if linked_signal_id ever pointed cross-team. New helper: src/test/fixtures/fake-supabase-fail-select.ts (NOT an edit to the shared createFakeSupabase fixture) — makes one table's SELECT chain fail with a given message, generalizing the ad hoc monkey-patched-`fake.from` idiom already used in golf's recurring-events.test.ts, needed to prove a sub-read failure is distinguishable from a genuinely empty table. All 6 files are plain `*.test.ts` (not `*.contract.test.ts`) so they run in vitest's `unit` project, not the 4-file `business` CI lane, per the spec's explicit naming risk. Gates: typecheck clean (tsc --noEmit, 0 errors repo-wide); eslint --max-warnings 0 clean on all 7 touched files; `vitest run --project unit src/contracts/baseball/` — 14 files, 93 tests, all passing (43 from the 6 new files + all 8 pre-existing files unaffected). No product-code changes — every assertion pins CURRENT behavior; no product-truth violation was discovered while writing these. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
#481) (#817) * fix(baseball): account for FairwayBottomNav clearance in Messages viewport math (#481) MessagesFairway's SHELL and ConversationClient's two height calcs only subtracted the AppShell top bar (4rem), never the fixed bottom-tab-bar clearance AppShell's content wrapper adds as extra padding-bottom whenever `bottomNav` is present. Since that ancestor is `min-h-dvh` (grows, not clips), the document became taller than the viewport and the composer's bottom portion rendered under the opaque FairwayBottomNav at the default scroll position on phones. Fix: subtract the existing global `--golf-mobile-bottom-nav-offset` var (globals.css `:root`; already the repo's shared idiom for this exact clearance — see KeyboardShortcutHint.tsx / InsightBulkActions.tsx / the golf round-review page) from both surfaces' height budgets. It resolves to `56px + env(safe-area-inset-bottom)` below `md` and `0px` at `md`+, i.e. it already matches AppShell's own bottomNav padding branch term-for-term and zeroes automatically once the bar stops rendering — no AppShell.tsx change needed. Also fixed: ConversationClient's two `100dvh-4rem` calcs were additionally missing the safe-area-inset-top term MessagesFairway's SHELL already had. Rewrote messages/[id]/loading.tsx's Suspense fallback (still on legacy warm-200/cream-50 tokens and the bare `100dvh-4rem` calc) to match ConversationClient's actual Living-Annual chrome and the fixed height formula. Scrubbed MessagesFairway.tsx's stale doc-comment claiming an `isRedesignEnabled()` page fork gates this component — the flag is hardcoded true (Wave W1) and MessagesClient.tsx always renders this shell unconditionally. Deviation from spec: the spec's approach proposed exposing a NEW CSS var on AppShell.tsx (shared with golf). A repo-wide grep found `--golf-mobile-bottom-nav-offset` already exists globally in globals.css for exactly this clearance, already consumed by golf components — so no AppShell.tsx or BaseballFairwayShell.tsx (frozen) edit was needed at all. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): account for coach HubSubNav height in Messages viewport (#481) Review found the prior fix still clipped the composer behind FairwayBottomNav for coaches: on /baseball/dashboard/messages(/[id]), resolveActiveHub() mounts a real HubSubNav ("Messages · Announcements") above these panes for coach role, but the height formulas never subtracted its height, so the pane's top (and thus its bottom edge) was pushed ~45px into the bottom nav's fixed region. HubSubNav (hub-sub-nav.tsx) now measures its own rendered height via a ResizeObserver-backed callback ref and publishes it to a new global CSS var, --baseball-hub-subnav-offset (0px default in globals.css, reset to 0px on unmount so subnav-less routes never inherit a stale value). MessagesFairway.tsx, ConversationClient.tsx (both branches), and messages/[id]/loading.tsx all subtract this var alongside the existing top-bar/safe-area/bottom-nav terms, so the box's own bottom edge clears FairwayBottomNav regardless of whether a sub-nav strip is mounted above it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…nk nav (#482) (#818) * fix(baseball): compact public player-profile mobile hero, fix back-link nav (#482) The public player profile's mobile hero was a tall (208px) decorative banner with absolute controls that ignored safe-area insets, a back control hardcoded to /baseball/dashboard/discover regardless of who opened the shared link, and two competing coach CTAs (Watchlist + Message) crammed into the overlay at once. - Hero banner: h-52 md:h-64 -> h-28 sm:h-36 md:h-64 (112px on phones, under the ~120px acceptance bar; desktop unchanged). - Passport-card overlap and mobile avatar re-tuned to match (-mt-28 -> -mt-10; w-32 h-32 -> w-20 h-20 avatar; verification badge scaled down proportionally) so identity clears the fold at 390x844 without colliding with the shrunk hero's controls. - Back control and coach-action bar now use `max(1rem, calc(env(safe-area-inset-top) + 0.5rem))` instead of a fixed `top-4`, clearing notch/Dynamic Island devices. - Back control is a history-aware button (mirrors BackChevron in src/components/ui/page-header.tsx): router.back() when history exists, else a contextual fallback (Discover for coaches, Player Today for the self-view case, home otherwise) instead of a single hardcoded Discover link. - Coach actions collapse to one primary CTA (Watchlist toggle) on <md; Message stays a full CTA at md+ and tucks into a small overflow menu on phones (dismissed via Escape or blur-outside). Deviation from the spec's suggested approach: used a small hand-rolled overflow menu (blur/Escape-dismiss, ink/paper styling) instead of src/components/ui/dropdown-menu.tsx — that component is unused anywhere in baseball and carries the cream/warm/primary token set from a different visual language than this page's paper/ink Living-Annual surface; a plain paper-card menu keeps the page's existing "one dark surface, rest is paper" rule intact. Coordination: left the dead recruiting-interests query in page.tsx (~lines 119-133) untouched per the other in-flight task that owns it. Gates: - npm run typecheck: clean, no output. - npx eslint --max-warnings 0 on the touched file: clean (0 problems) after adding tabIndex={-1} to the role="menu" panel and two precedented eslint-disable-next-line comments (helm/no-raw-button for the file's already-established raw-button idiom; jsx-a11y/no-static- element-interactions for the blur-only dismiss wrapper, matching the same disable pattern already used across src/app/golf/admin/crm/components/*). - No existing test file imports PlayerProfileClient.tsx (confirmed via repo-wide grep) and the spec names no test file to add; ran the adjacent src/lib/baseball/__tests__/public-profile-anon-gating.test.ts as a sanity check (80/80 passed, unaffected by this change). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(baseball): clear coach-action row on notched-iPhone worst case (#818 review) The mobile hero-height/passport-overlap pairing (h-28/112px hero, -mt-10/40px overlap) collided with the safe-area fix on real notched/Dynamic-Island iPhones: env(safe-area-inset-top) is a live value (viewportFit:'cover' in layout.tsx), and at its ~59px worst case (14/15/16 Pro Dynamic Island, portrait) the coach action row's bottom edge landed at ~111px while the opaque passport card's top edge stayed a fixed 72px — a 24-39px band where the card visually covered and intercepted taps on the Watchlist toggle / overflow-menu trigger. Re-derive the pairing so the card clears the row even at that worst case: mobile hero h-28 -> h-32 (128px), overlap -mt-10 -> -mt-2 (8px), giving the passport card a top edge of 120px against a worst-case row bottom of 111px (9px of real clearance). sm:/md: breakpoints are untouched (explicit sm:-mt-10 preserves the prior sm-range look now that the base value changed). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
… (#822) * test(baseball): add pgTAP suite for scope_player_ids staff isolation (#406) can_view_baseball_player() (20260630180000_baseball_scope_player_ids_rls.sql) had no behavioral test proving its scope_player_ids allowlist semantics. Add supabase/tests/rls/baseball_scope_player_ids_isolation.sql, mirroring the seed-then-impersonate pattern already proven in baseball_recalc_body_guards.sql / golf_coach_insights_cross_tenant_select.sql: seeds one team + one scoped (non-head, non-primary, active) staff coach, impersonates via SET LOCAL request.jwt.claims, and asserts empty/NULL scope (fallthrough visibility), a one-player allowlist, a multi-player allowlist, and a suspended-status short-circuit ahead of the scope check. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * test(baseball): add cross-staff isolation case to scope_player_ids pgTAP suite Review flagged that the "isolation" suite only ever seeded one staff-coach row per team, so a regression dropping/weakening the coach_id predicate in can_view_baseball_player()'s row lookup (WHERE tcs.team_id = p_team_id AND tcs.coach_id = v_coach_id LIMIT 1) would be structurally invisible to it. Adds Case D: a second staff-coach row on the same team with a wider scope_player_ids allowlist, then asserts (impersonating the first coach) that its results are unaffected by the second row's existence or scope. plan(13) -> plan(15) for the two new assertions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ests (#379 Phase 0) (#827) * fix(baseball): reconcile seed stats with Stats Center + add drift/smoke tests (#379 Phase 0) scripts/seed-baseball-stats.mjs now writes BOTH stats layers for every stat_type:'game' session it seeds: the legacy flat/aggregate rows (unchanged, for grandfathered consumers) AND matching baseball_box_score_batting/_pitching rows for a synthetic completed game, followed by the same recalculate_baseball_season_stats RPC games.ts's box-score save flow calls. Every write is an upsert (game id / (game_id, player_id) keys) — never delete-then-reinsert. seed-baseball-demo.ts's stale doc comment (claiming Stats Center should stay empty) now points at its stats-seeding companion script instead. Adds the two tests #379's acceptance criteria ask for: - seeded-stats-non-empty.smoke.test.ts runs the seed script's own seedTeamStats() against a fake Supabase client and asserts getStatsCenter() is non-empty for a "seeded" team. - command-center-stats-center-drift.test.ts asserts Command Center's legacy-layer game average and Stats Center's box-score-layer average agree for the same player/team/season. Adds manifest entries for the two new test files (they reference the deprecated table names as fixture/doc data) so the existing stat-layer contract test stays green, and updates stats-architecture.md's status notes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): review fixes for #379 Phase 0 — shared game rows + honest drift test mustFix #1: buildBoxScoreRowsForSessions minted a PRIVATE baseball_games row per player per session (id baked in player_id), so seeding a normal roster inserted 150-350 fabricated single-player "completed" games with no scores — flooding Stats Center's game counts, the Decision Room game-results rail, and the Postgame review game-picker. Fixed by introducing a shared team-level game schedule (buildTeamGameSchedule) that every player's game-type sessions attach to, and regrouping buildBoxScoreRowsForSessions/seedTeamStats to derive ONE shared, upserted game row per distinct team+date (never player-scoped), with every attending player's batting/pitching lines referencing that same game_id — mirroring the existing correct pattern in scripts/seed-rini-baseball-demo.ts. Verified empirically: 20 players now produce 23 shared team games (not ~150-350 private ones), each player appearing in ~9-13 of them. mustFix #2: the drift test's "passing" case hand-built a zero-practice- session baseball_player_aggregates fixture that bypassed seedTeamStats() entirely, overclaiming that Command Center's blended average and Stats Center's game-only average reconcile for realistic (~40% practice) players — they don't; that's a later #379 phase. Rewrote the test to run the REAL seedTeamStats() path with a deterministic (mulberry32-seeded) realistic game+practice mix and pin the honest picture: (1) the GAME-CONTEXT numbers genuinely agree between the two layers (no drift in the underlying game data — the actual Phase 0 reconciliation), and (2) Command Center's blended average still legitimately diverges from Stats Center's game-only average for a realistic player, asserted explicitly as an open Phase 0 gap rather than masked by a cherry-picked fixture. Updated docs/baseball/ stats-architecture.md's Phase 0 status note and the stat-layer-manifest.ts note to match. Gates: npm run typecheck (clean), eslint --max-warnings 0 on all touched files (clean), vitest run on both #379 test files + stat-layer-contract.test (11 files / 25 tests passed), plus the full src/contracts/baseball suite (52 files / 273 tests passed, no regressions) and the seed script's node --test safety suite (3/3 passed, still upsert-only/no .delete()). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* feat(baseball): add shared legacy-flat stat adapter module (#379) Generalizes roster-aggregates-merge.ts's box-score-over-legacy merge into src/lib/baseball/read-models/legacy-stat-adapters.ts — the one shared precedence rule (box-score > legacy-fallback > no-data, permanent practice carve-out, null-safe event-derived fields, sourceLayer provenance tag) that every later stats-layer-reconciliation migration will build on. roster-aggregates-merge.ts becomes a thin wrapper delegating to the new module; its exported mergeSeasonStatsIntoAggregates signature and behavior are unchanged so roster.ts and RosterClient.tsx need no changes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): restore per-field legacy fallback for rate stats in roster merge Review of #828 flagged that toLegacyAggregateShape() switched from a per-field legacy fallback to taking the box-score row's rate fields verbatim, dropping the pre-#379 `row.avg ?? existing?.career_avg ?? null` behavior for career_avg/career_obp/career_slg/career_ops/game_avg/ season_avg. Any pure pitcher/DH gets a baseball_player_season_stats row with g=0 and every rate field null (recalculate_baseball_season_stats nulls them when v_ab/v_pa = 0), which silently blanked their real, existing legacy figures to em-dash on the live Roster Wall. Restore the per-field fallback for exactly those six fields, mirroring the avg/max_pitch_velocity handling already in this file: box-score truth (including a legitimate 0) still wins outright whenever present; the legacy value is only read back in when the box-score-derived field is null/undefined. Add a test case for the pure-pitcher/DH shape (legacy row has real values, season-stats row has g=0 and null rates) plus a companion case proving a genuine box-score 0 still wins over a nonzero legacy figure, encoding the exact precedence. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ess lonely 1-tab hub strips (#829) Overnight mobile UI/UX audit — shared-shell group + primitive groundwork requested by other in-flight fixer agents: - PopoverPanel.Item: 36px -> 44px tap-target floor (min-h-11), app-wide. - Modal (ui/modal.tsx): new opt-in `sheetOnMobile` prop — bottom sheet (rounded-t, drag bar, safe-area padding, dvh-safe max-height) below `md`, centered dialog unchanged at md+. Absent prop renders byte-identical. - Fairway Sheet: new opt-in `mobileSide` prop — resolves to `mobileSide` below `md`, `side` at md+, so a desktop docked `side="right"` panel can become a bottom sheet on phone without every consumer hand-rolling its own useMediaQuery + ternary (the pattern FairwayNewMessageSheet/ InsightPanel already duplicate by hand). - Segmented: internal horizontal scroll + scroll-fade (useScrollFade) so a segment row wider than its container scrolls instead of silently clipping under the app's mobile overflow-x:clip guard. No-op when content already fits (fadeStyle is `{}`). - BaseballHelm HubSubNav / resolveActiveHub: suppress the sub-nav strip entirely when a hub resolves to < 2 visible tabs (was only 0) — kills the pointless single-tab sticky row on /baseball/dashboard/my-stats and /academics. Updated the one test that asserted the old single-tab resolve behavior. - FairwayHubSubNav (golf): 44px tap-target floor via an invisible vertical hit-slop pseudo-element, NOT a visible size change — golf's sub-nav row height (and LargeTitleContext's offset math) stays pixel-identical at every breakpoint. - Cross-referencing doc comments at both z-index token definition sites (design-tokens.css --fw-z-*, tokens.css --z-*) flagging the two differently-numbered ladders that share tier names. Unification is intentionally NOT done here (see PR body follow-up). Gates: tsc --noEmit clean; eslint --max-warnings 0 on all touched files clean; vitest run on nav-manifest.test.ts, resolve-active-hub.test.ts, and every nav|hub-matching suite (60 files / 1556 tests) green. Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…830) * fix(mobile): shared-primitive tap-target/overflow/sheet fixes + suppress lonely 1-tab hub strips Overnight mobile UI/UX audit — shared-shell group + primitive groundwork requested by other in-flight fixer agents: - PopoverPanel.Item: 36px -> 44px tap-target floor (min-h-11), app-wide. - Modal (ui/modal.tsx): new opt-in `sheetOnMobile` prop — bottom sheet (rounded-t, drag bar, safe-area padding, dvh-safe max-height) below `md`, centered dialog unchanged at md+. Absent prop renders byte-identical. - Fairway Sheet: new opt-in `mobileSide` prop — resolves to `mobileSide` below `md`, `side` at md+, so a desktop docked `side="right"` panel can become a bottom sheet on phone without every consumer hand-rolling its own useMediaQuery + ternary (the pattern FairwayNewMessageSheet/ InsightPanel already duplicate by hand). - Segmented: internal horizontal scroll + scroll-fade (useScrollFade) so a segment row wider than its container scrolls instead of silently clipping under the app's mobile overflow-x:clip guard. No-op when content already fits (fadeStyle is `{}`). - BaseballHelm HubSubNav / resolveActiveHub: suppress the sub-nav strip entirely when a hub resolves to < 2 visible tabs (was only 0) — kills the pointless single-tab sticky row on /baseball/dashboard/my-stats and /academics. Updated the one test that asserted the old single-tab resolve behavior. - FairwayHubSubNav (golf): 44px tap-target floor via an invisible vertical hit-slop pseudo-element, NOT a visible size change — golf's sub-nav row height (and LargeTitleContext's offset math) stays pixel-identical at every breakpoint. - Cross-referencing doc comments at both z-index token definition sites (design-tokens.css --fw-z-*, tokens.css --z-*) flagging the two differently-numbered ladders that share tier names. Unification is intentionally NOT done here (see PR body follow-up). Gates: tsc --noEmit clean; eslint --max-warnings 0 on all touched files clean; vitest run on nav-manifest.test.ts, resolve-active-hub.test.ts, and every nav|hub-matching suite (60 files / 1556 tests) green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): stats-surfaces mobile tap targets + games header wrap Overnight mobile UI/UX audit — stats-surfaces group (3/3 findings fixed): - StatsCenterClient: the season stepper, Batting/Pitching + Official/All side toggles (shared SegmentedControl), Clear-filters button, and every position chip explicitly override Button's 44px min-h floor with min-h-0. Ported the Fairway Button/Segmented coarse-pointer pattern ([@media(pointer:coarse)]:min-h-[44px]) so touch always clears 44px while the compact ~32px chip look is preserved for mouse/desktop. - GamesList (+ its loading.tsx skeleton): header was an unconditional `flex items-center justify-between` packing the title against the season Select + Refresh + Add-Game cluster with no wrap, crowding the title into a ragged sliver on 320-390px phones. Changed to `flex flex-col gap-3 sm:flex-row sm:items-center sm:justify-between` so the control cluster drops to its own row below the title on phone, matching the loading skeleton so there's no chrome drift. - EditGameModal: opened as a centered desktop-style dialog on every viewport. Wired the new `sheetOnMobile` opt-in on the shared Modal (landed via the primitives PR, cherry-picked below) so it renders as a bottom sheet under `md` and an unchanged centered dialog at `md`+. Deferred (not this group / requires an off-limits shared file): - Select's outer wrapper (src/components/ui/select.tsx:160) hardcodes `w-full` on its own div rather than forwarding the caller's className, so the `w-28` passed from GamesList only constrains the inner trigger button. In practice the flex-shrink default already caps it to the remaining row width, and stacking the header (this commit) gives that row its own full-width line, so the visible symptom from the finding is resolved — but the underlying non-forwarding wrapper is a src/components/ui/** primitives-task file this task may not touch. Left as a follow-up for that task. This commit includes a cherry-pick of primitives PR #829's head (068899c) — not yet merged into batch/bbh-finish-0714 at task start — for the `Modal.sheetOnMobile` prop EditGameModal depends on. That commit will appear as a duplicate here until #829 merges and this branch rebases/squash-merges past it. Gates: tsc --noEmit clean; eslint --max-warnings 0 on all 4 touched files clean; vitest run on GamesList.record-summary.test.tsx + product-trust.contract.test.ts (7 files / 30 tests) green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(mobile): shared-primitive tap-target/overflow/sheet fixes + suppress lonely 1-tab hub strips
Overnight mobile UI/UX audit — shared-shell group + primitive groundwork
requested by other in-flight fixer agents:
- PopoverPanel.Item: 36px -> 44px tap-target floor (min-h-11), app-wide.
- Modal (ui/modal.tsx): new opt-in `sheetOnMobile` prop — bottom sheet
(rounded-t, drag bar, safe-area padding, dvh-safe max-height) below `md`,
centered dialog unchanged at md+. Absent prop renders byte-identical.
- Fairway Sheet: new opt-in `mobileSide` prop — resolves to `mobileSide`
below `md`, `side` at md+, so a desktop docked `side="right"` panel can
become a bottom sheet on phone without every consumer hand-rolling its
own useMediaQuery + ternary (the pattern FairwayNewMessageSheet/
InsightPanel already duplicate by hand).
- Segmented: internal horizontal scroll + scroll-fade (useScrollFade) so a
segment row wider than its container scrolls instead of silently
clipping under the app's mobile overflow-x:clip guard. No-op when
content already fits (fadeStyle is `{}`).
- BaseballHelm HubSubNav / resolveActiveHub: suppress the sub-nav strip
entirely when a hub resolves to < 2 visible tabs (was only 0) — kills
the pointless single-tab sticky row on /baseball/dashboard/my-stats and
/academics. Updated the one test that asserted the old single-tab
resolve behavior.
- FairwayHubSubNav (golf): 44px tap-target floor via an invisible vertical
hit-slop pseudo-element, NOT a visible size change — golf's sub-nav row
height (and LargeTitleContext's offset math) stays pixel-identical at
every breakpoint.
- Cross-referencing doc comments at both z-index token definition sites
(design-tokens.css --fw-z-*, tokens.css --z-*) flagging the two
differently-numbered ladders that share tier names. Unification is
intentionally NOT done here (see PR body follow-up).
Gates: tsc --noEmit clean; eslint --max-warnings 0 on all touched files
clean; vitest run on nav-manifest.test.ts, resolve-active-hub.test.ts, and
every nav|hub-matching suite (60 files / 1556 tests) green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa
* fix(baseball): mobile UI/UX pass — development group (6 findings)
Overnight mobile audit fixes for the "development" group (practice
planner, scrimmage builder, dev plans, compare/comparisons):
- practice/TimeRailBuilder + PracticePlannerClient (critical): the rail's
drag handle (~14px) and resize strip (h-2) are far under 44px and the
keyboard Arrow-key alternative can't be reached from a touchscreen (no
focus-driven soft keyboard). Added Start/Duration +/-5min steppers to
the Block-details card — the true tap-based equivalent, wired through
the same updateBlock path the rail uses. RAIL_STEP/MIN_DURATION/fmtClock
are now exported from TimeRailBuilder so both surfaces share one source
of truth instead of a second copy that can drift.
- features/player-comparison.tsx (critical): the footer's "Comparing N
players" label + Export PDF/Save Comparison buttons had no wrap/stack
fallback; Card's `overflow-clip` silently hid whichever button lost the
fight for space under ~390px. Footer now stacks (`flex-col` -> `sm:flex-row`)
with a wrapping button row.
- practice-planner/ScrimmageLineupBuilder.tsx (major): field-position slot
buttons and the bench/bullpen remove (X) button are the ONLY tap
mechanism (native HTML5 drag doesn't fire on touch) but both overrode
Button's 44px floor with `min-h-0`. Dropped the override on the slot
button; grew the remove button to a real 44x44 box rather than an
invisible hit-slop pseudo-element, since list rows sit only 4px apart
(`space-y-1`) and an oversized hit-slop would steal taps from the
neighboring row's remove button.
- coach/CreateDevPlanModal.tsx (major): hand-rolled centered dialog with
static `vh` sizing and a hard `grid-cols-2` roster picker — the Save/
Cancel footer could end up hidden behind the on-screen keyboard, and
player cells crammed avatar+name+position into ~120px. Switched to
bottom-anchored-on-phone + `dvh`-safe max-height (centered dialog
unchanged at `sm:`+), `grid-cols-1 sm:grid-cols-2` picker, and safe-area
footer padding. Kept the hand-rolled shell rather than re-platforming
onto the shared Modal's `sheetOnMobile` (can't visually verify a larger
rewrite without a browser in this environment) — noted in the PR body.
- dashboard/compare/CompareClient.tsx (major): the "Add Players" search
box and the selected-player chip row sat side-by-side with no stacking
breakpoint and no `min-w-0`, so search became unusable as soon as a 3rd
player was selected. Stacks below `sm`; search wrapper can now actually
shrink.
- features/saved-comparisons-list.tsx (minor): the one un-migrated sibling
of dev-plans/compare/practice — plain `Card` + hand-rolled empty state
instead of the Living Annual kit. Ported onto `PaperCard` +
`<EmptyIssue variant="generic" ink="pursuit" />` (matching
PracticePlannerClient's usage) and swapped `warm-*` tones for
`text-text-*`/`--hairline`/`grade-plus` ink tokens.
Cherry-picked primitives PR #829 (068899c) onto this branch since it
wasn't yet merged into batch/bbh-finish-0714 — the duplicate commit will
disappear on squash-merge.
Gates: npm run typecheck (clean), npx eslint --max-warnings 0 on all 7
touched files (clean), npx vitest run on CreateDevPlanModal.test.tsx +
dev-plans-coach-gating.test.ts + practice-validation.test.ts (92/92
passed, 12 test files via shared setup).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Fable Integrator <fable@helm.local>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…#837) - SignalDrillDown: opt into the Sheet primitive's new `mobileSide="bottom"` (PR #829) so the phone drill-down renders as a native bottom sheet instead of the desktop docked `side="right"` panel — matching Doctrine rule 4 already applied to FairwayNewMessageSheet/InsightPanel. No hand-rolled useMediaQuery needed; the primitive resolves it. - command-center/loading.tsx: KPI skeleton now emits StatStrip's real GRID_BASE[2] + GRID_COLS[4] classes byte-for-byte (2x2 grid below `sm`) instead of a single stacked column, so the loading -> loaded transition no longer visibly reflows on phone. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
… content (#840) - Sticky save bars in ProgramSettingsClient and recruiting-preferences used a bare `bottom-4`, which sat ~16px above the viewport edge -- squarely inside FairwayBottomNav's opaque 56px+safe-area footprint, so the only save affordance on these settings forms was invisible/ untappable on mobile. Both now offset by `--golf-mobile-bottom-nav-offset` (zeroes at md), the same idiom the Messages sticky footer uses. - MinimumStandards' enabled/disabled toggle was `<Button variant="primary" className="w-6 h-6 ...">` with no `size`, so Button's default md `min-h-[44px]` (a different twMerge conflict group than h-6/w-6) stayed applied, forcing a 44px-tall/24px-wide sliver with the checkmark clipped. Swapped to a plain native button (Button's ripple effect needs `overflow-hidden`, which would clip the 44px hit-slop pseudo-element) sized correctly at 24x24 with an invisible `before:-inset-3` hit-slop restoring the 44px tap-target floor without growing the visible box. - RecruitingWeightDistributor's per-metric description was gated entirely behind onMouseEnter/onMouseLeave -- unreachable on touch. Now always rendered. - ProgramSettingsClient's three brand color swatches were h-10 (40px); bumped to h-11 (44px). - recruiting-preferences' 50-state picker was one flat wall of pills; grouped into US Census 4-region clusters so it reads as composed content on a phone instead of a dumped tag-cloud. Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…p, green pills, urgency wrap (#831) * fix(baseball): mobile messages-hub audit — skeleton drift, subnav clip, green pills, urgency wrap Overnight mobile UI/UX audit, messages-hub group (6 findings): - messages/loading.tsx (critical): replaced the generic, non-responsive `<SkeletonMessages />` (hardcoded w-80 + flex-1 desktop shape, overflows a 320px phone by ~48px, mismatched chrome) with `<MessagesFairway loading .../>` — the exact fallback MessagesClient's own <Suspense> already uses, so there is exactly one skeleton shape for this route instead of two that can drift out of sync. - HubSubNav height accounting (critical): MessagesFairway.tsx and ConversationClient.tsx already carry the `--baseball-hub-subnav-offset` fix (#481/#817, pre-existing on this branch) that reserves the coach "Messages · Announcements" sub-nav strip's real height so the composer never clips behind the bottom tab bar. Closed the one remaining gap: ConversationClient's composer form was missing the `pb-[calc(1rem+env(safe-area-inset-bottom))] lg:pb-4` safe-area padding MessagesClient's own composer already has — added it to match. - NewMessageModal (major): search-result rows used `<Button variant= "primary">`, which unconditionally sets `bg-primary-600 text-white` — every unselected contact rendered as a solid green pill. Switched to `variant="ghost"` so no unconditional accent background survives; the existing hover/selected classNames are untouched. - UrgencyPicker (major): `grid-cols-4` with no responsive override crammed a dot + label + description into ~57px cells even at 430px, wrapping character-by-character. Dropped to `grid-cols-2 sm:grid-cols-4` so each phone cell has room to breathe. - Compose flows (minor): added `sheetOnMobile` (new opt-in prop from the primitives PR, cherry-picked below) to both NewMessageModal's and CreateAnnouncementFlow's `<Modal>` so they render as bottom sheets below `md` instead of centered shrunk-desktop dialogs; this also gets the dvh-safe max-height for free (sheetOnMobile's branch already uses `100dvh`, unlike the default centered-dialog branch). - ConversationClient vs MessagesClient dual-design (minor) — DEFERRED, see PR body: fixing this correctly (retire `/messages/[id]` and redirect deep links to `?conversation=`) would require touching route-manifest/ test-registry files outside this finding's listed files (nav-manifest. test.ts, baseball-route-inventory.ts) and deleting a route with its own deliberate, recently-#481-fixed height-parity work and dedicated loading skeleton — disproportionate risk for a minor finding within this PR's scope. Cherry-picked PR #829 (shared primitives, still open at time of this run) onto this branch for `Modal`'s `sheetOnMobile` prop — see that commit's own message for its full scope (PopoverPanel 44px floor, Sheet mobileSide, Segmented overflow-scroll, HubSubNav <2-tab suppression). Duplicate commit disappears on squash-merge once #829 lands on the base branch first. Gates: tsc --noEmit clean; eslint --max-warnings 0 on all 5 touched files clean; vitest run on nav-manifest.test.ts, resolve-active-hub.test.ts, and use-messages.test.tsx (12 files / 640 tests) green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): composer sm-breakpoint safe-area padding gets reset by py-5 shorthand Review found the sm:640-1023px range dropped the safe-area bottom padding added in the prior commit: sm:py-5's own padding-bottom sub-declaration (1.25rem flat) wins over the base pb-[calc(1rem+env(safe-area-inset-bottom))] utility for any viewport >= 640px, since sm:'s media-query block is emitted after the unprefixed base rules. lg:pb-4 already masked this for desktop, so the flattened padding was only visible in the sm-to-lg gap (phone landscape / small tablet). Add sm:pb-[calc(1.25rem+env(safe-area-inset-bottom))] alongside the existing sm:px-6 sm:py-5 so the safe-area term survives through that range too, keeping lg:pb-4 as the final flat desktop value. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ets, honest roster skeleton (#832) * fix(baseball): mobile team-ops — touch-visible doc actions, 44px targets, honest roster skeleton Group: team-ops (mobile audit). Fixes all 8 findings (1 critical, 3 major, 4 minor): - DocumentCard.tsx (critical): kebab "More options" trigger + footer Preview/Download row were opacity-0 group-hover only — permanently invisible on touch. Now touch-visible by default, hover-gated only at md+; footer Download link gets an invisible hit-slop pseudo-element (IconButton's overflow-hidden ripple layer rules out hit-slop there, so the Preview IconButton instead drops its w-auto/h-auto override back to the shared 44px default). - ScrimmageLineupBuilder.tsx (major): removed min-h-0 overrides on the Team Blue/White toggle and roster-row select buttons (both restore the Button primitive's 44px floor); bench/bullpen remove (X) button gets an explicit min-h-11/min-w-11 since it was p-0 with no other floor. - PracticeEffectivenessClient.tsx (major): removed min-h-0 from the All/Measured/Attention/Need-data filter tabs, restoring the 44px floor without the horizontal padding change (kept the simpler "or" branch of the fixSketch over swapping to Segmented — lower risk, same outcome). - roster/loading.tsx (major): added a `md:hidden` stacked skeleton (avatar/name + 2 stat pills + row-menu box, no min-width/overflow-x) mirroring RosterFairway's real RosterWall mobile branch; existing 680px table skeleton now `hidden md:block`, so there's no more desktop-shaped skeleton flashing on a phone before data loads. - documents-client.tsx + TravelClient.tsx (minor): swapped both window.confirm(...) delete flows for the shared ConfirmDialog (danger variant, native bottom-sheet on-device), matching RosterMemberActions/SavedLineupsPanel elsewhere in the hub. - TravelClient.tsx (minor): handleSaved now calls router.refresh() instead of window.location.reload(). Fix-sketch deviation: TravelClient copies its itineraries prop into local useState, which does NOT re-seed on a bare router.refresh() (useState's initializer only runs once) — added a useEffect that re-syncs local state whenever the server-refreshed prop changes, so the created/edited/deleted trip actually appears instead of silently going stale. - PopoverPanel.Item 44px-floor finding: already resolved by the primitives PR (#829, cherry-picked below) — min-h-11 confirmed in PopoverPanel.tsx; no additional edit needed in RosterMemberActions.tsx. - PracticeRecapPanel.tsx (minor): scrimmage scoring grid is now `grid-cols-1 sm:grid-cols-3` so Blue/White/Innings stack on a 320px phone instead of squeezing into ~80px columns. Cherry-picked primitives PR #829 (head 068899c) onto this branch first — tap-target floors, Modal sheetOnMobile, Sheet responsive side, Segmented overflow. That commit will disappear as a duplicate on squash-merge of this PR against the primitives PR's own merge. Gates: npm run typecheck (clean) · npx eslint --max-warnings 0 on all 7 touched files (clean) · npx vitest run PracticeRecapPanel.test.tsx (4/4 passed, only touched-file test importer found). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(baseball): PR #832 review — wire real Travel refetch, fix Download hit-slop overlap Two mustFix items from mobile review verdict: - TravelClient.tsx: handleSaved() called router.refresh(), a no-op since TravelPageClient fetches itineraries entirely client-side. Lifted a lean reloadItineraries() callback (getTeamItineraries only — deliberately not the heavier detectRoleAndLoad, which would flash <PageLoading /> over the whole page) from TravelPageClient down as onReload, and handleSaved now calls it directly. The existing local-state resync effect is no longer dead: it's what reflects the refetched prop into TravelClient's mirrored state once onReload() resolves. - DocumentCard.tsx: the footer Download link's before:-inset-2.5 hit-slop (needed to reach the 44px floor on a ~26px visible box) expanded 10px left, but gap-1 between it and the Preview IconButton was only 4px — 6px of invisible hit area bled onto Preview's box and could steal taps meant for it. Widened the footer row to gap-3 (12px), clearing the 10px expansion with margin to spare; left the already-correct -inset-2.5 alone since shrinking it would drop Download's own target under 44px. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…r sheet, camp roster, map default (#833) * fix(baseball): mobile recruiting hub — peek panel/compare bar off-screen, filter drawer, camp roster row, map default Overnight mobile UI/UX audit — recruiting group (Discover/Pipeline/ Watchlist/Camps). Fixes every critical + major finding, plus one minor that lived in an already-touched file. - PeekPanelRoot (Pipeline/Discover/Watchlist player+team peek): was a hard-coded 512px right-edge panel with no responsive override — most of it rendered off the left edge of any phone. Now a bottom sheet below `sm`, right-edge dock at `sm:` and up. - CompareBar (Discover): unconstrained fixed-center row overflowed the viewport with no way to reach Compare/Clear, and sat inside the mobile bottom-nav's band. Now constrained + horizontally scrollable, and raised above the bottom nav. - DiscoverClient: replaced the hand-rolled left-edge drawer (no focus trap/Escape/scrim-click) with the app's own vaul Sheet — the codebase's own documented "no left drawer" rule. - FilterPanel: added a `sticky` opt-out (default true) so the mobile sheet instance doesn't fight the sheet's own header for the scroll container's top edge. - Discover Map default: Map is now the untouched default only at `md:`+; below `md` a first-time/returning visit lands on List (a full US choropleth shrinks small states below any tap target on phone). Map stays one tap away — picking it from ViewToggle this session still works. Also reduced the map card's own padding on phone. - PlayerCard (Discover grid): quick actions (watchlist/message) were hover-only with zero touch fallback — added the same `[@media(pointer:coarse)]` + `group-focus-within` treatment `<HoverReveal>` applies elsewhere (kept as direct classes since this card's `group` is shared by other children — see inline comment). - DiscoverView pagination pills: `size="icon-sm"` instead of a bare `w-8 h-8` fighting the default Button padding, clearing 44px. - CampDetailClient: roster row now stacks identity above the badge/Check-In row below `sm` (was one non-wrapping line where Check In was the first thing squeezed off); RosterFilterControl's `min-h-0` override dropped so it clears 44px. - CreateCampModal: rebuilt on the shared `Modal` (`sheetOnMobile`) — focus trap, Escape, scrim-click, dvh-safe height, safe-area footer padding, all for free instead of re-derived by hand. Cherry-picked PR #829 (shared-primitive groundwork: Modal `sheetOnMobile`, Sheet `mobileSide`, PopoverPanel.Item 44px floor, Segmented overflow scroll) as a prerequisite — not yet merged into batch/bbh-finish-0714 at task start. That commit disappears from this branch's history on squash-merge of #829. Deferred (minor, live in files not otherwise touched this pass): - CampsClient.tsx: coach-actions row `flex-wrap` at 320px. - PipelineClient.tsx/WatchlistClient.tsx: commit-ceremony toast overlaps HubSubNav. - ScoutPacketManager.tsx: share-link icon buttons under 44px. Gates: tsc --noEmit clean; eslint --max-warnings 0 on all 8 touched files clean; vitest run on discover/camp/peek-panel action tests (16 files, 68 tests) — all passing (no dedicated unit tests import these UI components directly). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(baseball): mobile review follow-up — pagination overflow + sheet overlay teardown Two mustFix items from PR #833 mobile review: - DiscoverView pagination row: clamp+scroll it exactly like CompareBar's own mobile fix in this same PR (max-w-[calc(100vw-2rem)] + overflow-x-auto + shrink-0 on Previous/pills-wrapper/Next), so the icon-sm 44px pill fix can't be undone by flex-shrink at narrow widths, and any overflow stays reachable instead of being clipped by the app's global mobile `overflow-x: clip` guard on html/body. - DiscoverClient mobile filter Sheet: stop gating visibility with className="lg:hidden" (only hides Drawer.Content — vaul's Drawer.Overlay is wired to `open` alone and ignores className, so a sheet left open while crossing lg+ left an invisible full-screen backdrop eating clicks). Force-close reactively via useMediaQuery + useEffect instead, mirroring the isDesktop pattern already used in the sibling DiscoverView.tsx, so the sheet and its overlay tear down together at the breakpoint. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…m/program/group split panes (#835) * fix(mobile): repoint Lift Lab session nav, clear bottom-nav-covered CTAs, phone-ify Lift Lab split panes performance-lift group — mobile UI/UX overnight pass (7 criticals, 3 majors, 1 minor). Criticals: - PlayerLiftSessionClient had no basePath prop (unlike its sibling PlayerLiftHomeClient) and hardcoded three '/lifting/dashboard/...' hrefs (BackLink, readiness-gate, post-completion). A BaseballHelm player tapping any of them on /baseball/dashboard/lift/[sessionId] was routed into the sibling Lifting Lab product. Added `basePath`, threaded from both pages ('/baseball/dashboard' and '/lifting/dashboard'), mirroring PlayerLiftHomeClient's contract exactly. Added a basePath link-wiring regression test (mirrors PlayerLiftHomeClient.test.tsx). - Same file: sticky "Complete lift" bar had no bottom offset / z-index, rendering under FairwayBottomNav. Offset via the existing --golf-mobile-bottom-nav-offset var, z-[var(--fw-z-sticky)]. - Same file: Card's own default padding + CardContent's own padding double-stacked (call-site fix only, no ui/card.tsx edit: padding="none" on the outer Card + explicit p-4/md:p-6). Rebuilt the per-set row (label + Log button on one line, reps/load/RPE as a 3-col grid) so it no longer needs a single unwrapped ~300px+ row; removed the min-h-0 tap-target overrides on the three inputs so they fall back to Input's 48px floor. - LiveWeightRoomClient: swapped the permanent w-80 athlete-detail `<aside>` for the shared Fairway `Sheet` (side="right" mobileSide="bottom") so it no longer competes with the athlete grid for width below 430px; swapped `h-screen` for a shell-aware `--golf-mobile-header-offset` / `--golf-mobile-bottom-nav-offset` height calc (same idiom as #481's ConversationClient fix) instead of assuming it owns the full viewport. - ProgramEditorClient / StrengthGroupsClient: both were hardcoded desktop split panes (permanent w-72/w-64 rail next to the primary content, no breakpoint). Added the same mobile tab-switcher LiftCanvas already uses one level up in this vertical (segmented Weeks&Days/Day-editor and Groups/Members tabs, `lg:` breakpoint to match LiftCanvas); wrapped the now-full-width prescriptions/members tables in overflow-x-auto so the pane's own content doesn't just overflow instead. Skeletons updated to match. - LiftBuilderClient: sticky Save bar had z-10 (below the nav's z-20) and a bare safe-area offset that didn't clear the nav's 56px height. Same --golf-mobile-bottom-nav-offset + z-[var(--fw-z-sticky)] fix. Majors: - PerformanceCommandCenter: dropped the min-w-[560px] + overflow-x-auto wrapper around WeightRoomRow — that row already stacks itself (flex-col ... sm:flex-row); the wrapper was forcing needless horizontal scroll on every phone for a problem the row's own layout had solved. - LiftCanvas: block reorder controls were `absolute -left-8` unconditionally, pushed off-screen below the 16px mobile gutter. Kept that treatment `lg:flex`-only; SessionBlock now renders its own inline up/down chevrons in the block header, `lg:hidden`. - ProgramListClient: header actions row (w-60 search + New program CTA) had no wrap; input now shrinks via flex-1/min-w-0 with the CTA pinned via shrink-0, flex-wrap as the last-resort fallback. Minor (fixed — cheap given the shipped primitives): - ExerciseWizard: added `sheetOnMobile` to its Modal call so the 7-step wizard is a bottom sheet below md instead of a shrunk centered dialog — the primitives PR built this exact prop for this exact case. Deferred: the LiftBuilderClient/LiftCanvas DetailPane Living-Annual re-skin (minor) — a visual redesign task carries real risk without browser verification (NO browsers/Playwright rail), so left for a follow-up with visual review rather than rushed here. Cherry-picked primitives PR #829 (068899c) onto this branch since it wasn't merged into batch/bbh-finish-0714 yet — that commit disappears on squash-merge of #829. Gates: tsc --noEmit clean; eslint --max-warnings 0 on all 13 touched files clean; vitest run on the new/touched test + every test importing a touched component (20 files / 96 tests) green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(mobile): review r2 — vaul-safe Sheet mount, restore desktop docked aside, subnav-aware height Addresses mustFix findings from the mobile-review-verdicts pass on #835: - LiveWeightRoomClient.tsx: the athlete-detail Sheet is now mounted unconditionally below `md` (visibility driven by `open`, body content gated on `selectedAthlete` inside `Sheet.Body`) so vaul plays its close/exit transition instead of the panel vanishing instantly — mirrors FairwayEventDetailDrawer.tsx's established pattern. At `md`+, restored the pre-existing non-modal docked `<aside>` instead of the Sheet (which is always a vaul `modal` Drawer.Root regardless of side/mobileSide) so a coach on desktop/tablet can still click a different athlete card without first closing the current detail panel. - LiveWeightRoomClient.tsx: SHELL_AWARE_HEIGHT now also subtracts `var(--baseball-hub-subnav-offset,0px)` as a third term, matching ConversationClient.tsx's established idiom, so the component under-fits correctly on /baseball/dashboard/performance/live (HubSubNav mounted) as well as its /lifting/dashboard/sessions/live sibling. - LabShell.tsx: scopes `--golf-mobile-bottom-nav-offset: 0px` for its own subtree (Lifting Lab never renders FairwayBottomNav), so PlayerLiftSessionClient's sticky "Complete lift" bar no longer floats above a phantom gap on /lifting/dashboard/lift/[sessionId] while BaseballHelm's route (BaseballFairwayShell, which DOES mount the nav) keeps the real non-zero offset. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…list-row fill (#836) * fix(mobile): import-center + stats-upload tap targets, chrome drift, list-row fill Overnight mobile UI/UX audit — import-center group. - StatsUploadClient/loading.tsx: drop the redundant min-h-dvh on the outer wrapper (AppShell's own <main className="flex-1"> already provides full available height); double-claiming 100dvh on top of the shell's own sticky chrome left 150px+ of dead scroll space below every wizard step. - StatsUploadClient: Match Players suggestion rows were <Button variant= "primary"> with only a hover override, so every unselected row rendered as a solid brand-green fill at rest. Swapped to variant="ghost" so the row's own border/hover classes are the only styling, matching the sibling toggle button one level up which already uses ghost correctly. - StatsUploadClient: Session Type toggle row had no flex-wrap, so the three labels were at real risk of spilling past the card edge at 320px. Added flex-wrap so it degrades gracefully instead of depending on exact fit. - ImportWizardClient: "change data shape" icon-only control was a hand-rolled rounded-full p-1 button around an 11x11 svg (~19x19px hit box). Routed through the shared fairway IconButton (variant="ghost" size="sm"), which already expands 36->44px under pointer:coarse — same primitive the rest of the kit uses for icon-only affordances. - ImportWizardClient: two secondary disclosure controls in the Preview/ Validate step (IssueGroup's blocking/warning/info header row, and DuplicateSummary's "Show what changes" text toggle) were under the 44px touch floor (~38px and ~18px respectively). Both are intentionally raw <button>s (documented reasons: the header row must not let <Button>'s variant fight the severity InkBadge; the text toggle must not take on <Button>'s padding/variant inside a dense summary row) — kept that, just added min-h-11 (+ inline-flex items-center on the text toggle) so the tap target clears 44px without changing either control's visual footprint. Deferred (explicit product decision, not a code fix): the /stats/upload vs /baseball/dashboard/import dual-wizard finding — Nick's call whether to retire /stats/upload in favor of Import Center or re-skin it onto the same Living Annual primitives. Left both wizards in place per task instructions. Gates: tsc --noEmit clean; eslint --max-warnings 0 on all 3 touched files clean; no existing vitest suite imports StatsUploadClient/ImportWizardClient (confirmed via targeted vitest run — "No test files found"), so no test gate applicable. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(mobile): gate import-center touch-target growth to coarse pointers Review found two of the #836 44px-floor fixes were applying unconditionally instead of only under `[@media(pointer:coarse)]:`, inflating desktop/tablet layout that shares a flex line with other content: - DuplicateSummary "Show what changes" toggle: min-h-11 was unconditional, resetting the row height even on mouse pointers where it shares a line with the "N new / N overwrite / N duplicate" summary spans. Now `[@media(pointer:coarse)]:min-h-11`. - "Change data shape" control: swapping in shared IconButton grew it to an unconditional 36px (size="sm") beside an ~18px InkBadge stamp on desktop/tablet — a visual change the PR body never called out. Reverted to the original raw pressableClass button, with touch-target growth to 44px confined to `[@media(pointer:coarse)]:h-11/w-11` only. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ces (#838) * fix(baseball): mobile-compose public team/program/player/packet surfaces Mobile UI/UX pass over the public (unauthenticated) share-link surfaces — team profile, program profile, player passport, scout packet. - team/[id] + program/[id]: header row (logo + Masthead/name) now stacks flex-col below sm instead of squeezing a 40-48px logo + text-5xl/text-3xl title into an ~88-158px column at 320-430px, mirroring the existing PlayerProfileClient flex-col md:flex-row hero pattern. Fixed in both the live page and its loading.tsx skeleton so mount doesn't jump layout. - player/[id]: PR #818 already collapsed the coach-viewing "Message" button into an overflow menu below md, but the "Add to Watchlist" pill was left full-label, so it still overlapped the absolute back button by ~28px at 320px. It now drops to an icon-only 44px control (aria-labelled) below md, matching the Message/overflow-menu breakpoint already in this file. - ScoutPacketCsvButton: dropped a stray `min-h-0` that was winning the tailwind-merge conflict against the Button component's own `min-h-[44px]` floor, collapsing the packet's one download control to ~32px tall. - OverviewTab "Physical & Metrics" grid: matched the single-column-on-mobile treatment StatsTab already uses for the same velocity/time fields, instead of a 2-col layout that risks numeral/unit collision at 320-430px. - "Schools of Interest" rows wore pressableClass (hover/press/focus-ring affordance) on a plain non-interactive div. They now link out to the school's program profile, same as the identity-chip links elsewhere on this page; a row with no organization record stays a plain div. - All four public-route error.tsx boundaries pointed "Go Home" at /baseball/dashboard, an authenticated route that immediately bounces anonymous share-link visitors to /login. Repointed to /baseball, which role-routes signed-in users and sends everyone else to login directly. Deferred (not in this PR): - "/baseball marketing root" finding — redirect-only landing is flagged as a product decision for Nick, per task instructions; left as-is. - ProgramRoster.tsx inline pitch/exit-velo stat blocks (minor, /baseball/ program/[id]) — lives in a file this PR doesn't otherwise touch; deferred per the "fix minors only in files already being edited" rule. Gates: npm run typecheck (clean), npx eslint --max-warnings 0 on all 10 touched files (clean), npx vitest run on the two source-scanning test suites that read team/[id] and program/[id] page.tsx (104/104 passed). No other test file imports PlayerProfileClient or ScoutPacketCsvButton. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(baseball): fill info-column width in mobile header flex-col mode Reviewer round 2 on PR #838: the team/program header info column used `min-w-0 flex-1` alone, which shrink-wraps to content width once the header switched to `flex-col` below `sm`. Both loading.tsx skeletons already pair `w-full` with `flex-1` for this exact column, so the live pages now match — `w-full min-w-0 flex-1`. No row-mode (sm+) regression: flex-basis:0% from flex-1 already wins over width there, same as the skeleton has always done. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ding, skeleton drift (#839) * fix(baseball): mobile player-experience tap-targets + skeleton drift (player-experience findings) - PlayerPassportFairway: batting stat grid base grid-cols-3 -> grid-cols-2 (matches the sibling Pitching block) so 2-3 digit tabular-nums numerals at the 40px clamp floor don't crowd/collide at 320-390px. - PlayerTimelineFairway: AckChip's Acknowledge/Acknowledged toggle gets an invisible before-pseudo-element hit-slop expanding its tap target to 44px without enlarging the visible chip (same technique as ModalShell/Sheet's close buttons). - PassportVisibilityControls: FieldRow's 4-option raw-button segmented control replaced with the shared Segmented primitive (44px under pointer:coarse, built-in overflow handling); its surface/border tokens inherit the Living Annual --paper/--hairline cream via the ambient .living-annual scope, so this also folds the panel's legacy cream-50/warm-200 chrome (exposure tiles, read-only banner, field-group container, footer divider) onto --paper-canvas/--hairline to match the rest of the passport page. - player/timeline/loading.tsx: skeleton rebuilt on fairwayScope('min-h-full bg-canvas') + --paper/--hairline PaperCard- shaped event rows, replacing the legacy bg-cream-100/bg-warm-100 skeleton that popped to the cooler canvas on load (same fix already applied to passport/today/practice loading.tsx). Gates: npm run typecheck (clean), npx eslint --max-warnings 0 on all 4 touched files (clean), npx vitest run src/components/baseball (40 files / 248 tests passed) — no test directly imports these presentation components, so the full baseball component suite stood in. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(baseball): AckChip hit-slop no longer crosses into title/time row Review on #839 flagged the symmetric before:-inset-y-3.5 hit-slop as crossing 4px into the title/time row above (10px mt-2.5 clearance - 14px top expansion). Made the inset asymmetric: before:-top-2 (8px, leaves 2px clear) / before:-bottom-3.5 (14px, unchanged, no neighbor below), keeping before:-inset-x-1.5 as-is. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(mobile): shrink AckChip top hit-slop to 6px — clears wrapped badge-line neighbor too Verifier-derived: before:-top-2 (8px) overshoots the 6px flex row-gap when the badge row wraps at 320-390px and AckChip lands on its own line; 6px clears both the 10px title-row case and the wrapped case. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
… fixes (#841) * fix(mobile): onboarding-auth chrome-drift skeletons + tap-target/grid fixes Fixes 5 major + 1 minor mobile findings in the onboarding-auth group (chrome that never got the same bespoke-skeleton/premium treatment as its login/signup siblings): - forgot-password, reset-password, complete-signup: loading.tsx now mirrors login/loading.tsx's narrow centered baseball-auth-field card (sized to each page's own form) instead of the generic dashboard-card PageLoading skeleton. - coach-onboarding: loading.tsx now mirrors the wizard's own EntryField + masthead + EditorialFrame double-bezel shell instead of PageLoading. - player onboarding: extracted PlayerOnboardingSkeleton (EntryField + HelmMark + a plain pulse card) reused as BOTH loading.tsx AND page.tsx's authLoading gate, collapsing three stacked, unrelated skeleton languages into one. - baseball/join/[code]: added a loading.tsx (previously none existed) mirroring the sibling staff/join/[code] flow's PaperCard skeleton — this page runs strictly more sequential Supabase awaits before its first paint, reached almost entirely via SMS/cellular invite links. - player onboarding "about"/"measurables" steps: NativeSelect 3-up grids (Bats/Throws/2nd Pos., Height ft/in/Weight) now stack to 2-up below `sm:` so each trigger keeps enough width for its option text to clear the chevron; select.tsx itself is off-limits tonight (primitives-owned). - player onboarding progress bar: swapped the shared, unconstrained StepIndicator dot-and-connector `<nav>` (no wrap/overflow guard, can push past 320px) for a new local StepProgress — a numbered-eyebrow + hairline rule, the same treatment coach-onboarding already uses. Kept as its own local copy (not imported from coach-onboarding's private _components folder) since that file documents itself as bespoke/non-shared. - login: "Forgot password?" link now has a 44px tap target via real padding (not a hit-slop pseudo-element) — it sits directly above the password Input with only a 6px gap, and an invisible ::before sized to 44px would have overlapped the Input's own tap area. Deviates from the literal fixSketch in two spots, both explained in the PR body: the tap-target technique (padding vs. hit-slop, to avoid stealing taps from the adjacent Input) and the StepIndicator swap (a fresh local StepProgress copy vs. importing coach-onboarding's private one). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(mobile): coach-onboarding loading.tsx matches 'type' step, not later chrome Review found the skeleton always rendered later-step-only chrome (a numbered progress rule + EditorialFrame double-bezel panel) even though the wizard's guaranteed first-rendered step ('type', gated by `step !== 'type'` in page.tsx) shows neither. Replaced the outer-bezel + 4 h-12 bars stand-in with an un-bezeled stack of 4 individually-bordered, OptionCard-shaped rows (rounded-2xl border p-5, icon swatch + two text lines) matching the real 'type' step's flat COACH_TYPES list. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…red legacy-stat adapter (#846) * feat(baseball): migrate Command Center game-context onto the #379 shared legacy-stat adapter Chunk "Migrate command-center.ts" of the #379 stats-layer reconciliation. rosterPulse[].careerAvg/totalSessions now resolve via legacy-stat-adapters.ts's adaptLegacyStatsMap: getStatsCenter()'s box-score-derived battingAll figures win whenever a player has box-score-era data this season, the raw legacy aggregate row is only a fallback tier (and the passthrough source for the still-unreplaced recentTrend field), and a player with neither is honestly no-data. This closes the Command-Center-vs-Stats-Center display divergence #827 documented for realistically-seeded players: both surfaces now show the identical game-context number by construction. - command-center.ts: fetch getStatsCenter() concurrently, reshape the legacy aggregate rows into the adapter's LegacyAggregateRow contract (missing columns explicit null, never fabricated), feed non-noData Stats Center rows in as BoxScoreGameContextRow, resolve the pulse via the adapter. - command-center-stats-center-drift.test.ts: the "KNOWN OPEN GAP" pin flips to RECONCILED — same real seedTeamStats() path, now asserting the displayed numbers agree while proving the raw legacy blend still differs (so the reconciliation is the adapter's doing, not a degenerate fixture). - command-center.test.ts: three new integration pins (box-score precedence, legacy-fallback no-regression, honest no-data) + stats-center table stubs for the existing timezone suite. - stat-layer-manifest.ts: command-center notes updated; entry retained because the adapter's legacy-fallback tier still requires a raw baseball_player_aggregates fetch here (documented in the note). Gates: npm run typecheck clean; eslint --max-warnings 0 clean on all four files; vitest 154/154 across the affected suites. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): stop pitching-only box-score rows from masking legacy career_avg in Command Center Command Center's boxScoreRows loop only skipped statsCenterModel rows where noData was true. A player with real box-score data for a non-batting side this season (pitching/catching/fielding/baserunning-only) has noData:false but battingAll.g === 0 / battingAll.avg === null — feeding that row into the #379 adapter as-is flipped sourceLayer to 'box-score' with a null avg, silently nulling out a real legacy career_avg fallback instead of falling through to it. Now skips on the games-count (battingAll.g === 0), not the derived rate, so a legitimate AB=0/g>0 batting row (e.g. walk-only game) still wins per the adapter's precedence. Also surfaces the adapter's sourceLayer on RosterPulseItem instead of discarding it, matching legacy-stat-adapters.ts's stated contract. Adds a regression test pinning a pitching-only-this-season + real legacy career_avg player to the legacy-fallback average, not a masked null. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
) * ci: promote BaseballHelm authed smoke to a required PR gate (#372) Adds a baseball-auth-smoke job to ci.yml that runs the already-built e2e/baseball-smoke.spec.ts + e2e/baseball-onboarding-smoke.spec.ts (with their existing fail-loud auth setup) as a required PR gate, folded into the `all` aggregate — the same promotion mechanism used for Supabase RLS tests (#517) and the import-cycle ratchet (#808). Previously this authenticated coach/player smoke only ran post-merge via playwright.yml's push/manual-only `e2e` job, which stayed untouched. Skips (not fails) on fork/Dependabot PRs, which get no repo secrets; same-repo pushes/PRs must have the required secrets or the job fails loudly. Docs updated to reflect the new hard-gate status and its added CI-minutes cost. Also unmasks the `|| echo "Playwright suite has failures..."` exit-code shim in playwright.yml's advisory "Playwright (chromium)" job so a real failure there turns the job red instead of silently passing (reviewer- flagged on #812 as blocking trust in the e2e gate). That job is advisory, not a required check, so this cannot newly block a merge — it only makes the status honest. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(ci): skip baseball-auth-smoke on Dependabot PRs, not just forks Dependabot PRs are opened against this repo (not a fork), so the existing head.repo.full_name == github.repository check alone did not exclude them. The job's own docs already claimed a Dependabot skip, but without secrets it would hard-fail on the pull_request branch, blocking the required `all` aggregate on every Dependabot PR. Add an explicit github.actor check and sync the "skips on fork/Dependabot PRs" prose in the job comment, branch-protection.md, and CI_RUNBOOK.md to match. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
#373) (#850) * test(e2e): authenticated Baseball route crawler, replace broken script (#373) scripts/route-crawler-baseball.mjs POSTed credentials to a /api/auth/login REST endpoint that never existed in this repo (BaseballHelm auth is a client-side Supabase form, not a JSON login API), so its sign-in always failed and it exited 0 ("no credentials — skipping") regardless of whether CI secrets were configured — and it was never wired into any workflow. Its "contract test" only regex-checked the source text, so it could never catch this. Replaces both with e2e/baseball-route-crawler.spec.ts, which runs under the existing baseball-coach/baseball-player Playwright projects (reusing the storageState auth baseball-smoke.spec.ts already established — no new login code) and discovers routes from the live rendered DOM (visible <nav> links, both the main sidebar and any hub-subnav strip) instead of statically parsing nav-registry.ts source. e2e/helpers/route-health.ts (shared with baseball-smoke.spec.ts, which now imports its ERROR_BOUNDARY_TEXT_RE instead of duplicating the regex) asserts each discovered route isn't a 4xx/5xx, doesn't bounce to /login, doesn't redirect into /golf/, doesn't render an error boundary, doesn't get stuck on a loading spinner, and isn't near-blank. Best-effort discovers public player/team/program/packet links surfaced on an authenticated page (capped at 3, only real linked routes, never guessed IDs) and re-verifies each in a fresh unauthenticated context. Wired into playwright.yml's e2e job as its own step + artifact upload. Deliberately NOT folded into #372's new required ci.yml gate: DOM-driven discovery and the stuck-spinner/near-blank heuristics are new, unproven surface area, so it runs advisory-only until it's demonstrated stable across several main-branch runs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(e2e): wire route-crawler spec into CI review mustFix items (#373) Review flagged three CI-wiring gaps that would make the new advisory crawler step deterministically fail on every main-branch run: - playwright.config.ts: baseball-coach/baseball-player projects only matched baseball-smoke.spec.ts, so the crawler step's --project filters resolved zero tests ("No tests found"). - playwright.config.ts: the bare chromium project's testIgnore didn't exclude the new spec, so it would also run unauthenticated there. - .github/workflows/playwright.yml: the crawler step lacked continue-on-error, so any future genuine failure/flake would cascade into skipping the subsequent required "Run Playwright tests" step, contradicting the step's own "advisory, doesn't block" comment. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ak baseline through shared adapter (#847) * fix(baseball): #379 Phase 2 — reconcile operational-signals cold-streak baseline through shared adapter Server actions Phase 2 chunk of the #379 stats-layer reconciliation design. - operational-signals.ts: loadPlayerHittingSpans's season/career OPS baseline for the player_cold_streak rule was reading baseball_player_stats WHERE stat_type='season' — a value no writer in this codebase has ever produced (imports.ts / stats.ts's uploadStatsCSV only write 'practice' | 'game' | 'other'). That lookup always returned [], so the cold-streak rule could never fire in production for any team. Fixed by sourcing the baseline through the shared legacy-flat adapter (legacy-stat-adapters.ts, #828), preferring the canonical box-score-era baseball_player_season_stats roll-up and falling back to the legacy baseball_player_aggregates row's career OPS only when no box-score-era row exists yet — the same precedence roster.ts/roster-aggregates-merge.ts already use for the same class of gap. The "recent N games" rolling window stays a direct baseball_player_stats read (no canonical per-game rolling-window primitive exists yet); manifest note updated to reflect the narrower remaining scope. - insights.ts: no code change needed — #819 (tonight's withBaseballAction migration) already deleted its only deprecated-table references as part of removing dead generateTeamInsights/getTeamInsights. Its GRANDFATHERED_CONSUMERS entry was stale (the stat-layer contract test's "no stale entries" check was red before this commit); removed here. - stats.ts: recalculatePlayerAggregates deliberately left as the raw legacy-row writer (documented why, not code-changed) — roster.ts, command-center.ts, and the player-today/-snapshot-cards/-passport read-models are still direct, unmigrated consumers of baseball_player_aggregates per the manifest, so retiring or bypassing this write path now would regress those surfaces to empty state for legacy-CSV-only teams. Matches the design's own stated alternative ("or is retired once nothing else needs a materialized baseball_player_aggregates row"). - New test: operational-signals-cold-streak.test.ts exercises the real runOperationalSignalDetection wrapped action (loadPlayerHittingSpans itself is intentionally not exported — every export of a 'use server' file is a real client-invokable action, and the observability-coverage contract enforces every such export is wrapped) across all three precedence tiers: box-score preferred, legacy fallback, no-data (rule does not fire). Gates: tsc --noEmit clean; eslint --max-warnings 0 on all touched files clean; targeted vitest run green (new test 3/3, stat-layer contract's stale-entry check now green, observability-coverage contract green, upload-stats-csv.test.ts unaffected/green). Pre-existing unrelated failures on this base branch (practice-effectiveness.test.ts, player-today-self-scope.test.ts, player-today-honest-loop.test.ts, engine-run-helm-lifting.test.ts + 3 others) confirmed present before this commit too (verified via git stash) — out of this chunk's scope, not touched. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa * fix(baseball): #379 PR review — per-field OPS legacy fallback + season/career honesty Addresses PR #847 review mustFix items: - operational-signals.ts's loadPlayerHittingSpans seasonOPSByPlayer build loop previously masked a real legacy career_ops whenever a box-score-era baseball_player_season_stats row existed for the player but its own OPS field was null (the pure-pitcher/DH shape: g=0 this season, avg/obp/slg/ops all null per recalculate_baseball_season_stats). Added a per-field fallback to legacyAggregates[playerId]?.career_ops, matching roster-aggregates-merge.ts's toLegacyAggregateShape precedent for the identical gap. - Added FactPlayerHittingSpan.seasonOPSSourceLayer (box-score/legacy-fallback/ no-data) so the player_cold_streak rule can label its baseline honestly: a legacy-fallback number is the aggregate row's LIFETIME career_ops, not a season-scoped figure, so evidence/whyItMatters text now says "career average" instead of an unqualified "season average" whenever the resolved value didn't come from a real box-score-era season row. - Added a 4th regression case to operational-signals-cold-streak.test.ts mirroring roster-aggregates-merge.test.ts's pure-pitcher/DH fixture: a season-stats row present with g=0 and all rate fields null, plus a legacy aggregates row with a real career_ops — asserts the rule still fires using the legacy baseline rather than silently producing 0 signals. Does not touch select.tsx (owned by #842). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa --------- Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…esh (#855) Two batch-level infra breaks blocking green CI on the assembled branch: 1. src/test/setup.tsx — ResizeObserver/IntersectionObserver mocks were vi.fn().mockImplementation(arrow); `new` constructs the arrow itself, which throws 'is not a constructor'. Latent since forever — first triggered by #829's use-scroll-fade (Segmented) constructing a real observer in any rendered test (FairwayRecruitingPage x8). Now real classes with vi.fn() method spies. 2. .cycles-baseline.json — check-cycles --update to lock in the 2 new benign madge-unresolvable externals (@supabase/supabase-js via #827's seed scripts, dotenv/config) per the script's own documented flow. Still 33 known cycles, none new. FairwayRecruitingPage.test.tsx: 9/9 passing after fix (was 8 failing). Co-authored-by: Fable Integrator <fable@helm.local> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
Too many files changed for review. ( Bypass the limit by tagging |
|
The latest updates on your projects. Learn more about Vercel for GitHub. 1 Skipped Deployment
|
|
Important Review skippedToo many files! This PR contains 202 files, which is 52 over the limit of 150. To get a review, narrow the scope: Upgrade to a paid plan to raise the limit. ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (202)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Updates to Preview Branch (batch/bbh-finish-0714) ↗︎
Tasks are run on every commit but only new migration files are pushed.
View logs for this Workflow Run ↗︎. |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| document.documentElement.style.setProperty(SUBNAV_OFFSET_VAR, `${node.offsetHeight}px`); | ||
| }; | ||
| publish(); | ||
| const ro = typeof ResizeObserver !== 'undefined' ? new ResizeObserver(publish) : null; |
njrini99-code
marked this pull request as ready for review
July 15, 2026 11:52
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
njrini99-code
enabled auto-merge (squash)
July 15, 2026 11:52
|
ⓘ Qodo reviews are paused because your trial has ended. Ask your workspace admin to add credits to resume reviews. Manage billing |
This was referenced Jul 28, 2026
Closed
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
The 2026-07-14→15 overnight mission batch: 45 PRs, each individually built by a Sonnet worker, adversarially reviewed (every finding either fixed+re-verified or documented), and merged into this integration branch. Every PR body carries its own gate evidence.
Assembled-branch gates (run on this exact tip, 66ec7ee)
Contents by lane
Deploy note
Merging this deploys prod (main auto-deploys). CI on this PR runs the full required set including the new hard gates.
Held for Nick (documented in the morning report)
Journey/pipeline vocabulary decision (memo ready) · #379 legacy backfill question · marketing root · dual upload-wizard consolidation · velocity eventDerived wiring follow-up.
🤖 Generated with Claude Code
https://claude.ai/code/session_01H9QAYqFTKsXGsVw6wXYssa