Phase 13 polish — product decisions #23

harshitsinghbhandari · 2026-05-30T15:20:35Z

harshitsinghbhandari
May 30, 2026
Maintainer

Tracking product-level decisions surfaced during Phase 13 polish work. Defaults applied unadjusted unless someone replies with an override.

Each top-level comment captures a single decision: question + proposed default + reasoning. Reply to the comment to override.

Phase 13 sub-tasks: a11y, keyboard, skeletons, errors, empty, mobile, dark mode, polish details.

PR: feat/phase-13-polish → develop (link added on PR open).

harshitsinghbhandari · 2026-05-30T15:40:08Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 — accessibility baseline implementer notes

A couple of product/design calls landed in this batch, flagging so the controller can sanity check:

Focus-ring color = --fg (i.e. solid black in light mode, solid white in dark mode), 2px outline with 2px offset. Picked over a brand-accent ring because we don't have a brand-accent token yet, and --fg guarantees ≥7:1 contrast against any surface in either theme. Easy to swap to a brand color later — it's a single CSS rule in globals.css.
--fg-subtle nudged from #737373 → #6b6b6b in light mode so the token clears 4.5:1 on both --bg (#fff) and --bg-subtle (#f5f5f5). Dark-mode #a3a3a3 already passed.
Optimistic-revert announcement copy. Engagement toggles now mount a hidden <span role="status" aria-live="polite"> that gets populated only when a revert happens. Copy I chose:
- LikeButton: "Like failed, reverted." / "Unlike failed, reverted."
- BookmarkButton: "Bookmark failed, reverted." / "Bookmark removal failed, reverted."
- FollowButton: "Follow failed, reverted." / "Unfollow failed, reverted."
  Short and direct — happy to swap to softer phrasing if it feels off.
PostCard title promoted from <h3> to <h2> so the home/latest/tag feeds don't skip from <h1> → <h3> (axe heading-order). Profile cards (ProfilePostCard) stay at <h3> because the profile page already has an <h2> section heading between the page <h1> and the cards.
Root layout no longer renders <main> — pages each render their own <main> (most already did). Sidesteps the duplicate-main landmark violation cleanly without forcing every page to share a single wrapper. Added <main> to the few stragglers (/auth/signin, /auth/blocked, /error, /not-found, /admin layout, /write, post page).
Enabled experimental.globalNotFound + added app/global-not-found.tsx so unmatched URLs still emit a complete document (Next 16 dev otherwise streams <html id="__next_error__"> with no lang attribute, which axe flags).

A11y spec result: 0 / 8 → 8 / 8 serious/critical-free public routes (/, /latest, /tags, /tag/agents, /search, /search?q=agent, /auth/signin, /auth/blocked). Typecheck, lint, and the 733 unit/integration tests are green.

0 replies

harshitsinghbhandari · 2026-05-30T16:10:15Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 — Keyboard navigation: DONE

Five logical commits on feat/phase-13-polish:

96680fa feat(phase-13): add skip-to-content link
3702253 feat(phase-13): add j/k/Enter feed traversal
4ef1f46 feat(phase-13): editor cmd/ctrl+enter publish + cmd/ctrl+s save
3704e79 feat(phase-13): esc + focus restore on modals
36bd743 test(phase-13): e2e coverage for keyboard nav

What shipped

/ focus search: already implemented in components/layout/NavSearch.tsx. Verified: skips <input>, <textarea>, <select>, and [contenteditable]. Modifier keys suppress (browser chords like Find-in-page still work).
j / k / Enter traversal: new <KeyboardFeedNav> (client) wraps each feed's <ul>. Cards opt in via data-feed-card + data-href + tabIndex={-1}. Wired into /, /latest, /tag/[slug], /bookmarks, and the profile PostList. Same input-guard as /.
Editor shortcuts: cmd/ctrl+Enter publishes (no-op when invalid or publishing); cmd/ctrl+s calls a new DraftManager.saveNow() imperative method that cancels the in-flight 30s autosave and flushes immediately. Listener mounted on window so it fires inside CodeMirror; ref-based to avoid re-attach on every keystroke.
Esc on modals:
- ReportModal: already had Esc; added focus restoration to the trigger element on unmount.
- DraftManager restore prompt: Esc dismisses without clearing the on-disk draft.
- DraftManager conflict prompt: Esc maps to "keep local" — non-destructive default.
- Note: delete-post and delete-comment "confirms" are window.confirm() (native browser dialog) — Esc / focus restoration are handled natively, no work needed.
Skip-to-content link: rendered as first child of <body> in app/layout.tsx; visible only on :focus/:focus-visible. Every page's <main> carries id="main-content" (audited 17 pages, all updated).

Verification

Check	Result
`rtk proxy pnpm typecheck`	pass
`rtk proxy pnpm lint`	0 errors, 15 pre-existing warnings
`rtk proxy pnpm test`	733/733 pass
`rtk proxy pnpm a11y`	8/8 pass
`rtk proxy pnpm e2e tests/e2e/keyboard.spec.ts`	3 pass, 1 skipped

The skipped e2e is the j/k/Enter traversal on /latest — it requires at least 2 cards to render, but the CI Supabase placeholder makes the feed empty (documented test.skip in the spec). The handler logic itself is structural and identical to what runs locally.

Product calls made

j/k vs arrow keys: went with j/k only (matches the brief and HN/Gmail vocabulary). Arrows can still be added later without breaking this.
Search guard scope: the existing NavSearch implementation already uses document.activeElement.tagName rather than e.target.matches(...). Kept it — it's semantically equivalent and one shape across both the / handler and the new <KeyboardFeedNav> handler.
Esc on draft-restore conflict modal: chose non-destructive "keep local" rather than "discard". Losing unsaved work to a stray Esc tap would be a much worse failure than an extra click later.
<KeyboardFeedNav> wrapping strategy: display: contents wrapper so it's layout-transparent for the existing UL/grid. Cards self-register via data-feed-card + data-href — no prop drilling.

Concerns

None blocking. The j/k spec's CI skip is a known limitation of the Supabase placeholder env, not a defect.

0 replies

harshitsinghbhandari · 2026-05-30T16:23:04Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 — Loading skeletons + Suspense (decisions)

Branch: feat/phase-13-polish · commits ebed0b7, 8a28ad6, d4bfdd9, 1d32592

Shimmer style: pulse (not wave)

Tailwind ships animate-pulse built-in; respects prefers-reduced-motion automatically.
Codebase already uses animate-pulse for the nav auth loading badge — keeping one motion vocabulary.
A gradient sweep ("shimmer wave") would need a custom keyframe + an extra utility class, and skeletons typically resolve in <300 ms so the cost isn't justified.

Default `count` for fallbacks

PostCardSkeleton — 5 (matches the typical viewport feed visible-card count without making the fold feel overwhelmed).
CommentSkeleton — 3 (most posts have a handful of comments; 3 conveys "list" without faking a wall of pending content).
SearchResultSkeleton — 5 (typical above-the-fold result density).
ProfileHeaderSkeleton — singular (profile headers are one-up by definition).
Each composite takes a count prop so callers can override per surface.

What got suspended

/ — <FeedList /> (For-You / Latest read + author/tag hydration)
/latest — <LatestList /> (paginated newest)
/tag/[slug] — <TagPostsList /> (post_tags lookup + posts query + hydration)
/search — <SearchResults /> (FTS RPC)
/[username] — <ProfileBody /> (pinned + authored)
/[username]/[type]/[slug] — <CommentsSection /> (thread walk)

Pages with a cursor/filter searchParam pass a <Suspense key=...> keyed off the params so changing filters or paginating re-shows the skeleton instead of holding the previous list.

What we deliberately did not suspend

Root layout.tsx (Nav, Footer, ThemeToggle, AuthProvider) — shared across all routes, expensive to flicker.
getSession() calls — cheap JWT decode, kept in the page shell so header copy ("For you" vs "Latest") can render with the static title.
Cached profile lookup (getCachedProfile, getCachedPost) — react.cache-wrapped one-shot queries.

Trade-off worth flagging

On /, the H1 now reads "For you" whenever the viewer is signed in, even if getForYouFeed errors and silently falls back to getLatestFeed inside the suspended body. Previously the heading was set to "Latest" after the fallback. The new behaviour avoids putting the session lookup inside Suspense and is acceptable since the fallback path is only hit on a transient DB error.

Verification (this run)

pnpm typecheck — green
pnpm lint — 0 errors (pre-existing warnings only)
pnpm test — 748 passed (85 files); added tests/unit/components/skeleton.test.tsx (12 assertions) and updated tests/unit/profile/page.test.tsx to traverse the new Suspense boundary
pnpm a11y — 8/8 passed

0 replies

harshitsinghbhandari · 2026-05-30T16:32:28Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 Section 4 — error boundaries shipped

Files added/changed

components/error/ErrorBoundary.tsx — minimal class-based primitive (~50 lines, no dep). Props: fallback, optional resetKey, children. Logs caught errors to console.error in componentDidCatch. resetKey change clears hasError so a parent can recover the boundary without a full page reload.
components/error/MdxFailedFallback.tsx — shared inline fallback (Couldn't render this <context>.) — copy-only, never renders the caught Error.
Six new route-level error.tsx files: [username], [username]/[type]/[slug], write, admin, search, tag/[slug]. Each is 'use client', renders title + short message + try-again button (calls reset()) + home link. Existing root app/error.tsx got the same "back to agentlab" link for consistency.
Risky widgets wrapped:
- <PostBody /> on the post page
- per-section in <StructuredSections /> (so one broken playbook/dive section can't kill its siblings)
- bio render in <ProfileHeader />
- <MermaidBlock /> in lib/mdx/components.tsx (PreWithMermaid override) — MermaidBlock already has internal try/catch on mermaid.render, but a sync error in dynamic import / state init would still bubble; the boundary now contains that

Decisions worth flagging for review

error.message / error.stack never rendered. Friendly copy only. The pinned test (error-route-fallback.test.tsx) passes a sentinel message and asserts it does NOT appear in the rendered output, so a future refactor that adds <p>{error.message}</p> lights up CI.
error.digest not surfaced either. Brief said digest is safe to render — I left it out for v1 because there's no support flow that asks users for it yet. Easy to add later as <p className="font-mono text-xs">Reference: {error.digest}</p> if/when we have a support inbox. Happy to add now if you'd prefer.
Tone is clinical, not jokey. "couldn't load this post" / "admin console hit an error" / "editor failed to load" — matches existing 404/500 copy. No "Oops!" / emoji.
reset vs unstable_retry. Next 16.2 added unstable_retry. The existing app/error.tsx uses reset, so I kept all six new ones on reset for consistency. Migration to unstable_retry is a one-liner per file if/when we want re-fetch semantics; for v1 the plain re-render is fine.
StructuredSections server-side throws still hit the route-level error page (the boundary is client-only). The per-section wrapper only catches hydration-time / client-render errors. That's the same trade-off as Suspense for async server components.
No new dep. react-error-boundary is ~20 lines of value, ~20 KB of cost; the in-tree class component is smaller than the import statement we'd have added.

Verification

pnpm typecheck — clean
pnpm lint — 0 errors (15 unchanged pre-existing warnings)
pnpm test — 757/757 passing (5 new tests added)
pnpm a11y — 8/8 still green

Commits on feat/phase-13-polish

feat(phase-13): add ErrorBoundary primitive for narrow widget recovery
feat(phase-13): add per-segment error.tsx with route-appropriate copy
feat(phase-13): wrap risky MDX/mermaid widgets in ErrorBoundary
test(phase-13): cover ErrorBoundary + post-page error route fallback

Not pushed, no PR yet — waiting for the broader Phase 13 PR per the brief.

0 replies

harshitsinghbhandari · 2026-05-30T16:39:26Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Task 6 — Empty states audit (commit `d6993df`)

Aligned every per-surface empty state with the Phase 13 brief.

Added / updated copy

/ (FeedList): "Follow people or wait while the feed warms up. Start with /tags." with a real <Link href="/tags">.
/admin/reports: "All caught up. No open reports."
/admin/tags: "No tags pending approval."
/<username> (PostList): "No posts yet." — covers both author-has-zero and filter-tab-empty cases.
/tag/<slug>: "No posts tagged here yet."

Verified-correct, no change

/bookmarks: "Bookmark posts to revisit them here." (Phase 8)
/search?q=...: "No posts match. Try fewer keywords." (Phase 9)
post page comments: "Be the first to comment." (Phase 7)

Tests added
tests/unit/pages/{home,admin-reports,admin-tags,profile,tag}.empty-state.test.tsx — each mocks the data layer to return zero rows and walks the Suspense child to assert the new copy renders. The home test also asserts the /tags action href.

Verification (rtk proxy)

pnpm typecheck clean
pnpm lint 0 errors (15 pre-existing warnings, unchanged)
pnpm test 765/765 (+5 new)
pnpm a11y 8/8

Deviations

Did not add a components/feedback/EmptyState.tsx shared component — each empty state is a single <p> styled by the surrounding page (home-feed__empty, profile-posts__empty, text-fg-subtle text-sm, etc.) and abstracting it would have added indirection without shared structure to lift. The existing per-surface CSS classes already drive the look.
Did not modify PinnedPosts (the brief mentioned per-tab profile state, but PinnedPosts returns null when empty, which is correct — only the authored list needs copy).

0 replies

harshitsinghbhandari · 2026-05-30T16:49:38Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Mobile breakpoints — decisions

Phase 13 Task 7 (mobile breakpoints) shipped in 5 commits on feat/phase-13-polish. Decisions worth recording:

Editor pane breakpoint: `lg` (1024px), not custom 900px

Brief target was ~900px for the editor split-pane to stack. Picked Tailwind default lg (1024px) instead of a custom breakpoint so the responsive vocabulary across the codebase stays consistent — every other surface uses sm: 640, md: 768, lg: 1024. The 124px deviation is acceptable; the editor still stacks cleanly at every viewport in the audit set (375/414/768/1024).

Editor tabs: simple Write/Preview tabs (not toolbar buttons)

Used role="tablist" with two <button role="tab"> elements that flip a view: edit | preview state. The tablist is lg:hidden, and each pane has matching aria-controls + lg:block to restore the side-by-side layout above the breakpoint. The drag-divider is hidden lg:block since it only makes sense in the desktop two-column mode.

Admin tables: dual-render, not single-source with CSS pivot

For /admin/tags and /admin/audit, kept the original <table> as-is and added a parallel <ul> of card-shaped <li> rendered with the same row data. Cards are md:hidden, table is hidden md:block. Duplication is fine at ≤6 columns; a single-source pivot via CSS grid-template would have made the table semantics murkier for the screen-reader path on desktop.

Profile pinned grid: explicit breakpoints

Previous grid-template-columns: repeat(auto-fill, minmax(260px, 1fr)) was responsive but non-deterministic across the audit viewports. Switched to explicit 1 / 2 / 3 cols at <640px / 640-1023px / 1024px+.

Post body: 70ch cap

.post-page already maxes at 720px; added max-width: 70ch to .post-body so paragraphs stay in the 50-75ch readability zone even when the article column reaches its 720px ceiling.

Nav search: already collapsed (Phase 9)

The CSS .nav-search__input already collapses to 2.5rem at <=640px via media query and expands on :focus-within. No code change needed; covered by the new mobile e2e spec.

Mobile e2e spec

New tests/e2e/mobile.spec.ts runs at 375x812 (iPhone 13/14 portrait — narrowest brief viewport). 8 public routes sweep + collapsed-nav check + 3 auth-gated routes + 4 admin routes. Auth-gated tests self-skip when E2E_TEST_AUTH_USER_ID / ADMIN_GITHUB_LOGINS_FOR_E2E are absent, so CI keeps the public sweep only.

Verification

typecheck: clean
lint: clean (pre-existing 15 unused-var warnings in test files unchanged)
vitest: 765/765 passing
a11y: 8/8 passing
mobile e2e: 9/9 ran (7 admin/auth-gated correctly skipped without env)

Commits

1ce3067 feat(phase-13): stack editor panes + add Write/Preview tabs on mobile
999d78c feat(phase-13): cap post body to 70ch for prose readability
a7f0089 feat(phase-13): pinned posts grid to explicit 1/2/3-col breakpoints
b362099 feat(phase-13): convert admin tables to card lists below md breakpoint
0ad4886 test(phase-13): mobile-viewport sweep at 375px

0 replies

harshitsinghbhandari · 2026-05-30T17:02:19Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 / Task 8: Dark Mode Audit — landed

Branch: feat/phase-13-polish. Four commits, all green (typecheck, lint, 769/769 vitest, 16/16 axe = 8 routes × 2 themes).

Surfaces touched

app/globals.css — added @custom-variant dark (&:where([data-theme='dark'], [data-theme='dark'] *)) so MDX components can express dark-mode counterparts inline via dark: utilities. Also added [data-theme='dark'] override for the lone non-token color (.comment-form__counter--over), light red shade for dark mode.
lib/mdx/components.tsx — Callout (info/tip/warning/danger) keeps the semantic hue with darker bg + lighter text in dark mode. Aside / Detail / Figure caption / Embed fallback / Pre swap neutral-* → theme tokens (bg-bg-subtle, text-fg-subtle, border-border).
lib/mdx/MermaidBlock.tsx + mermaid-init.ts — subscribed to data-theme mutations via useSyncExternalStore (same pattern as ThemeToggle). Toggling theme re-runs the render effect with the new mermaid theme so diagrams re-tint live. Error pill gets dark variant.
app/layout.tsx — pre-hydration inline <script> in <head> reads localStorage.getItem('theme') (fallback: matchMedia('(prefers-color-scheme: dark)')) and sets data-theme on <html> before React mounts. Eliminates FOUC. <html suppressHydrationWarning> because server can't know client's theme.
components/layout/ThemeToggle.tsx — toggle() now writes localStorage.setItem('theme', next), wrapped in try/catch for Safari private mode.
tests/unit/components/theme-toggle.test.tsx (new) — 4 tests: render label, light→dark, dark→light, localStorage-throws survival.
tests/e2e/theme.spec.ts (new) — toggle persists to localStorage; theme survives navigation.
tests/e2e/a11y.spec.ts — added a second pass per route with addInitScript forcing theme=dark so axe color-contrast catches dark-palette regressions (8 → 16 tests).

Decisions for the record

Mermaid theme strategy: dynamic. I went with the MutationObserver/useSyncExternalStore route rather than a single neutral theme, because mermaid's built-in default is light-only and dark is dark-only — there's no neutral variant that reads cleanly against both bg: #fff and bg: #000. The trade-off is that the diagram re-renders (not just re-styles) on theme flip, which causes a brief blank frame. Acceptable for a rare user action.
localStorage key: theme. Chose the short, conventional key over agentlab-theme. devtools-readable, matches the pattern used by next-themes and most blog templates. If we ever multi-tenant, we revisit.
No new color tokens. Reused existing --bg, --fg, --bg-subtle, --fg-subtle, --border, --bg-hover. Semantic callouts (info/tip/warning/danger) intentionally stay on Tailwind palette colors (blue-*, emerald-*, amber-*, red-*) — promoting these to tokens would be premature; they're MDX-author-controlled, not chrome.

Deferred / not done

Prism syntax token colors. No .token.keyword etc. styles exist anywhere — Prism tokens currently render monochrome via var(--fg) inheritance. Legible in both themes, just not colored. Promoting this to a styled prism theme is its own task (and probably belongs in Phase 14 once we add a real code-heavy post to test against).
OG image cleanup. Only /og.png exists; app/layout.tsx, app/[username]/[type]/[slug]/page.tsx, and app/[username]/page.tsx all reference exactly that one file. No light variant to drop — already correct.

Commits

feat(phase-13): wire dark variant + scrub hardcoded neutrals in MDX surfaces
feat(phase-13): reactive mermaid theme + dark-mode error fallback
feat(phase-13): persist theme choice across reloads + navigation
test(phase-13): cover theme persistence + axe dark-theme sweep

0 replies

harshitsinghbhandari · 2026-05-30T17:38:53Z

harshitsinghbhandari
May 30, 2026
Maintainer Author

Phase 13 polish — shake vs. toast (and title hierarchy)

Decision: shake.

Why shake (not toast)

Toast needs a global container, portal, and queue — significant new infra for a feedback path that already has an accessibility-correct sr-only aria-live equivalent (Phase 13 shipped that with a monotonic counter so consecutive identical reverts re-announce).
Shake is colocated CSS on the failing button itself, no new components.
Matches the well-worn macOS "wrong password" affordance — immediately legible.
prefers-reduced-motion: reduce zeroes the keyframes, so motion-averse users get the sr-only announcement only.

Implementation notes

@keyframes shake (~6px translate, 400ms, 4 cycles) + .shake-on-revert class in app/globals.css.
LikeButton / BookmarkButton / FollowButton: useEffect on the existing revert.n counter toggles the class via a ref + setTimeout(…, 400), then removes it.
Tried re-keying the button on revert.n first — would unmount/remount the element on every failure, steal keyboard focus, and lose disabled mid-flight. Direct DOM toggle avoids that.
Unit tests use vi.useFakeTimers() + vi.advanceTimersByTime(401) to assert the class is gone after the duration.

Side decision: title-template hierarchy

While auditing titles to “{label} — agentlab.in”, admin sub-pages caused a triple-dash problem. Resolution:

Root layout: title.template = “%s — agentlab.in”, default: “agentlab.in”.
Admin layout: nested title.template = “%s · Admin — agentlab.in”, with title.absolute: “Admin — agentlab.in” to bypass the root template at the admin index.
Convention: middle dot (·) for in-app hierarchy, em-dash (—) reserved exclusively for the site separator. Keeps the site-name separator unambiguous.
Dynamic pages (post / tag / profile / followers / following): title.absolute because they build their own “… — agentlab.in” string and don’t want a second suffix.

Leaving as a note here since the brief asked for any tagline conventions surfacing from the title audit to be posted.

— worker age-16, branch feat/phase-13-polish

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Phase 13 polish — product decisions #23

Uh oh!

{{title}}

Uh oh!

Replies: 8 comments

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Uh oh!

Phase 13 polish — product decisions #23

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer

Replies: 8 comments

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Phase 13 — Keyboard navigation: DONE

What shipped

Verification

Product calls made

Concerns

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Phase 13 — Loading skeletons + Suspense (decisions)

Shimmer style: pulse (not wave)

Default count for fallbacks

What got suspended

What we deliberately did not suspend

Trade-off worth flagging

Verification (this run)

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Phase 13 Section 4 — error boundaries shipped

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Task 6 — Empty states audit (commit d6993df)

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Mobile breakpoints — decisions

Editor pane breakpoint: lg (1024px), not custom 900px

Editor tabs: simple Write/Preview tabs (not toolbar buttons)

Admin tables: dual-render, not single-source with CSS pivot

Profile pinned grid: explicit breakpoints

Post body: 70ch cap

Nav search: already collapsed (Phase 9)

Mobile e2e spec

Verification

Commits

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Phase 13 / Task 8: Dark Mode Audit — landed

Surfaces touched

Decisions for the record

Deferred / not done

Commits

Uh oh!

harshitsinghbhandari May 30, 2026 Maintainer Author

Phase 13 polish — shake vs. toast (and title hierarchy)

Why shake (not toast)

Implementation notes

Side decision: title-template hierarchy

harshitsinghbhandari
May 30, 2026
Maintainer

harshitsinghbhandari
May 30, 2026
Maintainer Author

harshitsinghbhandari
May 30, 2026
Maintainer Author

harshitsinghbhandari
May 30, 2026
Maintainer Author

Default `count` for fallbacks

harshitsinghbhandari
May 30, 2026
Maintainer Author

harshitsinghbhandari
May 30, 2026
Maintainer Author

Task 6 — Empty states audit (commit `d6993df`)

harshitsinghbhandari
May 30, 2026
Maintainer Author

Editor pane breakpoint: `lg` (1024px), not custom 900px

harshitsinghbhandari
May 30, 2026
Maintainer Author

harshitsinghbhandari
May 30, 2026
Maintainer Author